Precision Medicine for personalized treatment of asthma

Petteri Teikari, PhD
https://www.linkedin.com/in/petteriteikari/
Version “Tue, March 26, 2024“
Precision Medicine
for personalized
treatment of
asthma
Endotypes and multimodal
deep learning for clinical practice,
academic research and clinical trials

About the slides
What are these?
Shallow literature analysis of asthma diagnostics and management, to aid the design
of new digital solutions
Format?
Even though these are slides, these are not meant as presentation aids. More like a
“visual literature review” without being as factually detailed as a “real review”.
Who are these for?
For engineers / data scientists developing “asthma hardware” or are involved with
precision asthma modeling
What are these not?
Not much about clinical workflows or realities that are blocking you from translating
your tech innovations to clinical practice

See separate docs for
●
ML and Signal Processing (“software part”)
●
Wearable Continuous Acoustic Lung Sensing (“hardware part”)

Asthma in a nutshell
Asthma is heterogeneous
Multiple endotypes exist requiring personalized endotype-specific treatment.
(van der Burg and Tufvesson (2023), Ray et al. (2022))
Over and underdiagnosis too common
Usually due to the lack of objective lung function testing which can demonstrate variable
expiratory airflow limitation that will support the diagnosis of asthma (Levy et al. (2023))
Effort-free lung function measures
Spirometry requires certain operator/patient skill. Non-invasive effort-free lung function measure
desired. Wearable impulse oscillometry type of measure (e.g. Samay Sylvee)?

Asthma
Umbrella term for various
asthma subtypes

Asthma Basics #1a
Levy et al. (2023): “Key recommendations for primary care
from the 2022 Global Initiative for Asthma (GINA) update”
The Global Initiative for Asthma (GINA) was established in 1993 by the
World Health Organization (WHO) and the US National Heart Lung and
Blood Institute to improve asthma awareness, prevention and
management worldwide. GINA develops and publishes evidence-
based, annually updated resources for clinicians. GINA guidance is
adopted by national asthma guidelines in many countries, adapted
to fit local healthcare systems, practices, and resource availability.
Asthma treatment is not “one size fits all”; GINA recommends
individualized assessment, adjustment, and review of treatment. As
many patients with difficult-to-treat or severe asthma are not referred
early for specialist review, we provide updated guidance for primary
care on diagnosis, further investigation, optimization and treatment of
severe asthma across secondary and tertiary care. While the GINA
strategy has global relevance, we recognize that there are special
considerations for its adoption in low- and middle-income countries,
particularly the current poor access to inhaled medications.
In order to ensure diagnosis of asthma is considered as early as possible,
clinicians should maintain a high index of suspicion when patients present
with respiratory symptoms12
.
Over- and under-diagnosis of asthma are common and are usually
due to the lack of objective lung function testing which can
demonstrate variable expiratory airflow limitation that will support the
diagnosis of asthma and help to exclude other causes13,14
. For continuity of
care, it is important to ensure that the diagnosis is recorded in each
patient’s medical record, detailing the basis for the diagnosis, including
objective measurements of variable airflow obstruction and airway
inflammation, if available. These details are often lacking in the medical
records of children15
and adults treated for asthma16,17
.
There is no single test for confirming the diagnosis of asthma. First, a
clinical diagnosis starts with a history of respiratory symptoms (such as
cough, wheeze, difficulty breathing and/or shortness of breath) that
typically vary over time and intensity. Symptoms of asthma are often
worse at night and in the early morning, and may be triggered by
factors such as viral infections, allergen exposure, exercise, strong smells,
cigarette smoke, exhaust fumes and laughter.

Asthma Basics #1b
Levy et al. (2023): “Key recommendations for primary care
from the 2022 Global Initiative for Asthma (GINA) update”
The GINA diagnostic flowchart 2022 Investigating poor symptom control and/or exacerbations
despite treatment

Asthma Basics #1c vs COPD
Levy et al. (2023): “Key recommendations for primary care from the 2022 Global Initiative for Asthma (GINA) update”

Asthma Basics #1* vs COPD
Bouwens et al. (2022): “Diagnostic differentiation
between asthma and COPD in primary care using
lung function testing”
Asthma and COPD are defined as different disease
entities, but in practice patients often show features of
both diseases making it challenging for primary care
clinicians to establish a correct diagnosis 5,6
. We aimed to
establish the added value of spirometry and more
advanced lung function measurements to differentiate
between asthma and COPD.
However, it is important to realise that the lung function
tests in our study were conducted by well-trained staff
in a pulmonary function laboratory and interpreted by
experienced chest physicians. To translate these results to
the real-life setting, it requires standardized procedures,
quality assurance and trained clinicians to interpret the
spirometry data accurately and this may be difficult to
achieve in primary care46,47,48
. Furthermore, the availability
of inflammatory markers in primary care could
potentially provide better discriminating diagnostic ability
but we did not investigate this in the current study.
Decision tree used by the chest physicians to support their assessment of
chronic lung disease diagnoses based on GOLD and GINA guidelines22
.

Asthma Basics #1e
Two-track options for
personalized management
of asthma for adults and
adolescents, to control
symptoms and minimize
future risk.

Asthma Basics #1f
Initial medications for adults and adolescents
diagnosed with asthma

Asthma Basics #2
NICE guideline [NG80] UK (March 2021): “Asthma: diagnosis, monitoring
and chronic asthma management”

Asthma Basics #3
https://www.asthmaandlung.org.uk/conditions/asthma/types-asthma
If you have asthma, knowing which type you have can
help you manage it better. Most types of asthma are
managed in the same way, with:
●
a preventer inhaler to use every day (even if you feel well)
●
a reliever inhaler to use when you have asthma symptoms
●
a written asthma action plan that includes information
about how your asthma affects you and how to manage it
●
regular asthma reviews with your GP or asthma nurse.
Evidence shows that sticking with these basics helps
people with asthma stay well.

Asthma in Numbers #1
asthmaandlung.org.uk:
Analysis from the BTS Difficult Asthma Registry in
2015 showed estimated annual treatment costs of
£2,912–£4,217 per patient; this is driven by
medication costs and, in those patients with
frequent exacerbations, hospital admissions.2
In
those with severe disease who are not receiving
high-dose treatment, the cost can be even higher.
In England and Wales alone, 1,302 people were
recorded as dying of asthma in 2015,3
while the
2014 National Review of Asthma Deaths,
commissioned by the Royal College of Physicians
(RCP), identified deficiencies in care of two-thirds of
the 195 cases of asthma mortality that they
reviewed.4
Chan et al. (2022)

Pessoa et al. (2022): “Automated respiratory
sound analysis”
UK has highest lung disease deaths in Western Europe

Crude asthma mortality rates during the bronchodilator and anti-inflammatory
eras. Reproduced from Pavord et al. (2018) - Larsson et al. (2020)
Given the availability of evidence-based reports and
guidelines, along with a range of effective medications
and inhaler devices to deliver those medications to the
target tissues, most patients nowadays should have
well-controlled asthma. Regrettably, however,
although hospital admissions and asthma mortality
have decreased over recent decades, rates now
appear to have plateaued2,7,9,30,31,32,33
. Several real-world
surveys have indicated that, at best, only 50% of
patients with asthma meet the criteria for well-
controlled asthma, indicating either that these criteria
are too strict or that asthma management is
inadequate12,28,29,34,35,36
.
Few patients are aware of the treatment goals outlined
in the guidelines38
. A UK-wide study showed that 58% of
patients were initially satisfied with the standard of
their asthma management and control38
. However,
after being shown international asthma guidelines on
the outcomes they should expect from their treatment,
this declined to only 33%38
.

Eric A Finkelstein et al. (2021): (Health Services and Systems
Research Program, Duke-NUS Medical School, Singapore;
AstraZeneca) “Economic burden of asthma in Singapore”
A total of 300 adults and 221 parents of children with asthma
were included in analysis. The total annual cost of adult asthma
was estimated to be SGD 1.74 billion (US$1.25 billion) with 42%
coming from the uncontrolled group, 45% from the partly
controlled group, and 13% from the well-controlled group. For
children, the total cost is SGD 0.35 billion (US$0.25 billion), with
64%, 26% and 10% coming from each group respectively.
Combined, the annual economic burden of asthma in Singapore
is SGD 2.09 billion (US$1.50 billion) with 79% due to productivity
losses.
Poorly controlled asthma imposes a significant economic
burden. Therefore, better control of disease has the potential to
generate not only health improvements, but also medical
expenditure savings and productivity gains.
Our results are in line with prior estimates. We find that 87% of
our sample has uncontrolled or partly controlled asthma, which
is identical to that reported in a prior study in Singapore. We
estimate annual asthma per capita healthcare costs across all
symptom groups at SGD 1290 (US$930).
Consistent with Singapore being a low-cost provider of health
services, this estimate is below that of Switzerland (SGD 2603),
USA (SGD 2021), and Hong Kong (SGD 1957) after accounting for
inflation and converting to SGD

Asthma an Umbrella Term
Howard et al. (2015): “Distinguishing Asthma Phenotypes
Using Machine Learning Approaches” Cited by 127
Asthma is not a single disease, but an umbrella term for a
number of distinct diseases, each of which are caused by a
distinct underlying pathophysiological mechanism. These
discrete disease entities are often labelled as 'asthma
endotypes'.
Scherzer et al. (2019): “Heterogeneity and the origins of
asthma”

T2 vs non-T2 type of asthma #1
EMJ 2018: Type 2 Inflammation and the Evolving Profile of Uncontrolled
Persistent Asthma
Type 2 asthma is so named because it is associated with
Type 2 inflammation and typically includes allergic asthma and
moderate-to-severe eosinophilic asthma, Prof Canonica explained. By
contrast, non-Type 2 asthma commonly has an older age of onset and
is often associated with obesity and neutrophilic inflammation.
Esteban-Gorgojo et al. (2018): Non-eosinophilic asthma: Current
perspectives Cited by 119

T2 vs non-T2 type of asthma #2
Wenzel (2012): “Asthma
phenotypes: the evolution from
clinical to molecular
approaches” Cited by 2894

Type2(-High) vs low-T2
Ray et al. (2022): “Determining asthma endotypes and outcomes:
Complementing existing clinical practice with modern machine learning”

Eosinophilic vs Non-Eosinophilic #1
Inflammation Biomarkers in the Assessment and Management of
Severe Asthma – Tools and Interpretation Recognition of severe
asthma phenotypes can inform the use of targeted therapies (LAMA
= long-acting muscarinic antagonists, LABA = long-acting beta
agonists)
https://www.severeasthma.org.au/biomarkers-recommendation/
Carr et al. (2018): “Eosinophilic and Noneosinophilic Asthma”
Clinical phenotypes, endotypes, and therapeutic options in
eosinophilic and noneosinophilic asthma.

Eosinophilic vs Non-Eosinophilic #2
Heaney et al. (2021):
“Eosinophilic and
Noneosinophilic Asthma: An
Expert Consensus Framework
to Characterize Phenotypes
in a Global Real-Life Severe
Asthma Cohort”

Allergic vs Nonallergic
https://doi.org/10.1038/nm.3300 allergic vs nonallergic eosinophilic airway inflammation
Schiffers et al. (2023): “Asthma
Prevalence and Phenotyping in the
General Population: The LEAD (Lung,
hEart, sociAl, boDy) Study”

Eosinophilic vs Neutrophilic
Kere et al. (2022): “Exploring proteomic plasma
biomarkers in eosinophilic and neutrophilic
asthma” Cited by 4
Few biomarkers identify eosinophilic and
neutrophilic asthma beyond cell concentrations in
blood or sputum. Finding novel biomarkers for
asthma endotypes could give insight about
disease mechanisms and guide tailored
treatment. Our aim was to investigate clinical
characteristics and inflammation-related plasma
proteins in relation to blood eosinophil and
neutrophil concentrations in subjects with and
without asthma.
Eosinophilic asthma was associated with a clear
clinical phenotype. With our definitions, we
identified MMP10 as a possible plasma biomarker
for eosinophilic asthma and CCL4 was linked to
neutrophilic asthma. These proteins should be
evaluated further in clinical settings and using
sputum granulocytes to define the asthma
endotypes.
http://dx.doi.org/10.1164/ajrccm.159.4.9708080

Symptom/Eosinophilic
Nissen et al. (2019): “Clinical
profile of predefined asthma
phenotypes in a large cohort
of UK primary care patients”

Adult vs Early Childhood
Ray et al. (2022): “Determining asthma endotypes and outcomes: Complementing existing clinical practice with modern
machine learning”

Symptom and Trigger-based phenotypes
Pérez de Llano et al. (2019): “Phenotype-Guided Asthma
Therapy: An Alternative Approach to Guidelines”
Proportion of adults with difficult-to-treat or severe asthma.
Levy et al. (2023)

Obesity-related asthma #1
Reyes-Angel et al. (2022): “Obesity-related asthma in children and adolescents”
McLoughlin and McDonald (2021): "The Management of
Extrapulmonary Comorbidities and Treatable Traits; Obesity,
Physical Inactivity, Anxiety, and Depression, in Adults With
Asthma"
Baffi et al. (2015):
“Asthma and
obesity: mechanisms
and clinical
implications”

Obesity-related asthma #2
View of the hormonal regulation of airway responsivenes state, adiponectin plays an
anti-inflammatory role, while leptin n (b) In obesity/insulin resistant condition, there is a
reduction in a levels, which in turn plays a pro-inflammatory role in the lungs airways.
Ach: acetylcholine; LR: leptin resistance. - Leiria et al. (2015): “Obesity and asthma:
beyond T(H)2 inflammation.”
Yu et al. (2023): “The relationship between the use
of GLP-1 receptor agonists and the incidence of
respiratory illness: a meta-analysis of randomized
controlled trials”
GLP-1 RAs may reduce asthma through
weight loss and improved insulin
sensitivity [Cardet et al. 2016, Rodrigues et al. 2017,
Rodrigues et al. 2020]
. In addition, in a study that
included 16,690 patients, patients with
GLP-1RA experienced less worsening of
chronic lower respiratory disease
compared to DPP-4I users [Albogami et al. 2021]
The treatment — semaglutide
(Ozempic®, Novo Nordisk) ... “Our trial
represents the first prospective study of
this class of medications in individuals
with asthma,”

Diagnostics room for improvement #1
Daines et al. (2020): “Defining high probability when making a
diagnosis of asthma in primary care: mixed-methods consensus
workshop”
Misdiagnosis of asthma is common with both underdiagnosis and
overdiagnosis reported.1–3
Asthma is difficult to diagnose because it is a
heterogeneous disease with different underlying disease processes, defined by
variable symptoms and expiratory airflow limitation.4
Investigations can determine
key features of asthma, but all have limitations.
For instance, spirometry is considered the reference standard for diagnosing
airway obstruction,5
but airway obstruction may not be persistent or present in all
cases of asthma.6
Bronchial provocation is the reference standard for determining bronchial hyper-
responsiveness7
but is time-consuming, carries a risk of inducing severe
bronchospasm8
, and is not widely available internationally.
Fractional exhaled nitric oxide (FeNO) identifies airway inflammation and is
useful for ruling in but less helpful for ruling out asthma.9
Peak flow variability
(PEF) is straightforward to perform but has low diagnostic accuracy for asthma.10 11
In the absence of an investigation that can rule in or rule out asthma with high
certainty, the diagnosis of asthma is often made clinically.
Chisholm et al. (2024): “Challenges in
diagnosing asthma in children”
●
Asthma can be diagnosed in children under 5, but is
unlikely to explain recurrent respiratory symptoms in
children under 2
●
Tests can be done to help support (or exclude) a clinical
diagnosis but should not be used solely to make (or
exclude) a diagnosis of asthma
Asthma is characterised by recurrent episodes of
cough and wheeze and difficulty in breathing. It affects
more than 10% of children in the UK, but no
universally agreed definition or diagnostic test
exists for asthma in children or adults. Diagnosis in
children mostly relies on a suggestive history and
response to a trial of preventer treatment. Lack of
objectivity in diagnosing asthma leads to both
overdiagnosis and underdiagnosis. Clinicians and
researchers have explored the role of objective tests,
such as spirometry, peak flow variability, and exhaled
nitric oxide in diagnosing asthma, and despite the lack
of evidence, several organisations have incorporated
these objective tests into diagnostic algorithms.

Diagnostics room for improvement #2
Daines et al. (2019): “Systematic review of clinical prediction
models to support the diagnosis of asthma in primary care”
Cited by 24
Diagnosing asthma is challenging. Misdiagnosis can lead to
untreated symptoms, incorrect treatment and avoidable deaths.
The best combination of clinical features and tests to achieve a
diagnosis of asthma is unclear. As asthma is usually diagnosed in
non-specialist settings, a clinical prediction model to aid the
assessment of the probability of asthma in primary care may
improve diagnostic accuracy.
We included seven modelling studies; all were at high risk of bias.
Model performance varied, and the area under the receiving
operating characteristic curve (AUROC) ranged from 0.61 to 0.82.
Patient-reported wheeze, symptom variability and history of allergy
or allergic rhinitis were associated with asthma. In conclusion,
clinical prediction models may support the diagnosis of asthma
in primary care, but existing models are at high risk of bias and
thus unreliable for informing practice. Future studies should
adhere to recognised standards, conduct model validation and
include a broader range of clinical data to derive a prediction
model of value for clinicians.
Forest plots demonstrating the strength of association of predictor variables
against the outcome asthma. Not all studies had extractable data. PP =
private practice, Co = combined dataset (private practice and primary care),
OPD = out-patient department, PC = primary care.

Spirometry #1
Gold standard for lung function, bulky and costly, requires patient effort
Feher (2017): “Lung Volumes and Airway Resistance”
Spirometers can measure three of four lung volumes, inspiratory
reserve volume, tidal volume, expiratory reserve volume, but
cannot measure residual volume.
Four lung capacities are also defined: inspiratory capacity, vital
capacity, functional residual capacity, and the total lung capacity.
Pulmonary ventilation is the product of tidal volume and
respiratory frequency. The maximum voluntary ventilation is the
maximum air that can be moved per minute. Spirometry also
provides a measure of airway resistance by use of the forced
expiratory volume test. The clinical spirogram presents the forced
vital capacity differently. In laminar flow, pressure necessary to
drive flow increases linearly with the flow. In turbulent flow, pressure
increases with the square of the flow. The Reynolds number is used
to estimate whether flow is laminar or turbulent. Airway resistance
also increases inversely with lung volume because stretch of the
lungs opens airways. Dynamic compression limits flow at high
expiratory effort.

Spirometry #2
Levy et al. (2009): “Diagnostic Spirometry in Primary Care: Proposed
standards for general practice compliant with American Thoracic
Society and European Respiratory Society recommendations.” Cited by 294

Spirometry #3
Doe et al. (2023): “Spirometry services in England post-
pandemic and the potential role of AI support software: a
qualitative study of challenges and opportunities”
Spirometry services to diagnose and monitor lung disease in primary
care were identified as a priority in the NHS Long Term Plan, and are
restarting post-COVID-19 pandemic in England; however, evidence
regarding best practice is limited.
Stakeholders highlighted historic challenges and the damaging
effects of the pandemic contributing to inequity in provision of
spirometry, which must be addressed. Overall, stakeholders were
positive about the potential of AI to support clinicians in quality
assessment and interpretation of spirometry. However, it was
evident that validation of the software must be sufficiently robust
for clinicians and healthcare commissioners to have trust in the
process.
This qualitative work forms the first stage of an evaluation of the use
of an AI decision support software (ArtiQ.Spiro, ArtiQ nv, Leuven,
Belgium) in the primary care respiratory diagnostic pathway (
https://www.artiq.eu, NIHR)
Home (low-cost) spirometry with AI decision support software coming? Can you assess lung function
with acoustic sensing? Complement spirometry with more frequent acoustic sensing?
Song et al. (2022): “SpiroSonic: Monitoring Human Lung
Function via Acoustic Sensing on Commodity
Smartphones” Cited by 63

Spirometry #4
Kocks et al. (2023) (General Practitioners Research
Institute, GPRI, Spiro@Home): “Feasibility, quality and
added value of unsupervised at-home spirometry in
primary care”
At-home spirometry could provide added value for
the diagnosis and monitoring of obstructive pulmonary
disease in primary care. However, it is unknown whether
implementation in a real-world setting is practicable
and produces good quality spirometry. We studied
feasibility, quality and added value of at-home
spirometry in primary care practices in the
Netherlands and Sweden.
Of 140 participants, 89.3% completed a home
spirometry session, of whom 59.2% produced
acceptable spirometry. Overall, HCPs and participants
rated home spirometry as feasible and of added value
for asthma and COPD monitoring in primary care,
though less helpful for diagnostic purposes. A small
mean difference in spirometry results was observed,
with FEV1 and FVC at-home being 0.076 L and 0.094 L
higher than at the GP office, respectively.
Bland-Altman plots of the agreement
between the best FEV1 (A) values from the
NuvoAir (Hernández et al. 2018) home session
and measured at the healthcare centre.

Peak Flow Meter (PEF)
Jain and Kavaru (1997): “A practical guide for peak expiratory flow
monitoring in asthma patients” Cited by 8
Peak flow monitoring has several limitations (TABLE 4) .
Perhaps the most serious of these is the potential for
inappropriate therapeutic action based on an inaccurate
peak flow record. A consistent trend of change in peak
flow is more convincing evidence for altering therapy than
an isolated finding. Because regular peak flow monitoring
and record-keeping requires considerable commitment,
patient compliance is a substantial problem.
Low-cost, but requires effort/skill from the patient leading to inaccurate measurements

Methacholine challenge test
American Lung Association
Also known as bronchoprovocation test is
performed to evaluate how "reactive" or
"responsive" the lungs are.
During the test, you will be asked to inhale
doses of methacholine, a drug that can
cause narrowing of the airways. A
breathing test will be repeated after each
dose of methacholine to measure the
degree of narrowing or constriction of
the airways.
The test starts with a very small dose of
methacholine and, depending on your
response, the doses will be increased
until either you experience 20 percent
drop in breathing ability, or you reach a
maximum dose with no change in your
lung function.
Kang et al. (2024): “Novel Artificial Intelligence-Based Technology to
Diagnose Asthma Using Methacholine Challenge Tests”
The methacholine challenge test (MCT) has high sensitivity but
relatively low specificity for asthma diagnosis. This study aimed to
develop and validate machine learning (ML) models to improve the
diagnostic performance of MCT for asthma.
Zachariades et al. (2024): “A new tidal breathing measurement device
detects bronchial obstruction during methacholine challenge test”
While spirometry relies on proper patients' cooperation and precise
execution of forced breathing maneuvers, we conducted a comparative
analysis with the portable nanomaterial-based sensing device,
SenseGuard™ (NanoVation-GS), to non-intrusively assess tidal
breathing parameters.
This study demonstrates that tidal breathing assessment using
SenseGuard™ device reliably detects clinically relevant changes of
respiratory parameter during the MCT. It effectively distinguishes
between responders and non-responders, with strong agreement to
conventional spirometry-measured FEV1.

FENO50 exhaled nitric oxide fraction #1
Miskoff et al. (2019): “Fractional Exhaled Nitric Oxide
Testing: Diagnostic Utility in Asthma, Chronic Obstructive
Pulmonary Disease, or Asthma-chronic Obstructive
Pulmonary Disease Overlap Syndrome”
Fractional exhaled nitric oxide (FeNO) is an
endogenous gaseous molecule which can be
measured in the human breath test because of
airway inflammation. It has been studied extensively
as a marker of inflammation and has been
incorporated into an algorithm for asthma
management.
https://doi.org/10.2147/JAA.S190489
https://doi.org/10.1007/978-94-007-6627-3_34

FENO50 exhaled nitric oxide fraction #2
Högman et al. (2024): “ERS technical standard: Global Lung
Function Initiative reference values for exhaled nitric oxide
fraction (FENO50)”
Elevated exhaled nitric oxide fraction at a flow rate of
50 mL·s−1 (FENO50) is an important indicator of T-helper
2-driven airway inflammation and may aid clinicians
in the diagnosis and monitoring of asthma. This
study aimed to derive Global Lung Function Initiative
reference equations and the upper limit of normal for
FENO50.
Available FENO50 data collected from different sites
using different protocols and devices were too
variable to develop a single all-age reference
equation. Further standardisation of FENO devices and
measurement are required before population
reference values might be derived.

Impulse Oscillometry #1
Meraz et al. (2013): “IOS is a multifrequency oscillation method; it provides measures of respiratory mechanics in
terms of respiratory impedance as a function of frequency Z (f). Respiratory Impedance is the transfer
function of pressure (P) and flow (Q). The respiratory Impedance (Z) measured by IOS is a complex quantity
and consists of a real part called respiratory Resistance (R) and an imaginary part called respiratory
Reactance (X). IOS also includes hallmarks such as Resonant Frequency (Fres) and Reactance Area (AX) also
known as the “Goldman Triangle”. IOS yields these indices over a selected frequency range of 3 to 35 Hz (
Goldman 2001).”
Promising “novel method’ requiring less patient effort, characterizes the response
(transfer function) of the respiratory system to a low-frequency (infra)-sound stimulus

Impulse Oscillometry #2
Smith (2019)
forced oscillation technique (FOT)
and impulse oscillometry (IOS) –
Brashier and Salvi (2015)
Mean a) AX and b)
Δ ΔX5 values in healthy subjects, and
patients with asthma, COPD and ILD.

Lung Wheezes
See the lung sound hardware doc for more lung auscultation literature
Kocks et al. (2023): “Development of a tool
to detect small airways dysfunction in
asthma clinical practice”
Small airways dysfunction (SAD) in
asthma is difficult to measure and a gold
standard is lacking. The aim of this study
was to develop a simple tool including items
of the Small Airways Dysfunction Tool (SADT)
questionnaire, basic patient characteristics
and respiratory tests available depending on
the clinical setting to predict SAD in asthma.
If access to respiratory tests is limited (e.g.
primary care in many countries), patients
with SAD could reasonably well be identified
by asking about wheezing at rest and a few
patient characteristics. In (advanced)
hospital settings patients with SAD could be
identified with considerably higher accuracy
using spirometry and oscillometry.
Noble and Donovan (2023): “If your
patient with asthma wheezes when
sitting or lying quietly, lung function
testing may reveal small airway
disease”
The paradigm of “small airway disease” (SAD)
continues to be advanced with considerable
enthusiasm by clinicians and scientists alike,
having full knowledge that the binary
classification of airways as “small” or
“large” is somewhat arbitrary.
Nonetheless, important questions that remain
unanswered are: 1) how does SAD manifest;
2) how do we detect SAD, ideally in a cost-
effective manner that can be adopted in
primary care; and finally 3) once a patient has
been classified as having SAD, what can we
do about it? The second of these questions is
amply addressed in the present issue of the
European Respiratory Journal by
Kocks et al. (2023)

Bronchodilator reversibility Spirometry
Ahmed et al. (2023): “Bronchodilator reversibility testing in morbidly obese non-
smokers: fluticasone/salmeterol efficacy versus salbutamol bronchodilator”
A positive response in reversibility testing is widely used to diagnose patients
with airway limitations. However, despite its simple procedure, it doesn’t accurately
reflect the exact airway irreversibility. This study aimed to investigate the efficacy of a
bronchodilation reversibility test using salbutamol and fluticasone/salmeterol
combination in obese non-smoker subjects.
Spirometry is pivotal in assessing bronchodilator reversibility. It is considered the
standard gold standard for diagnosing diseases with obstructive airways in general
practice. Reversibility testing measures the airflow expiration response after
inhaling a bronchodilator. A positive test is defined as an increase in FEV1 of more
than 12% from baseline or a 200-ml increase, according to the American Thoracic
Society (ATS). In comparison, the European Respiratory Society (ERS) recommended a
change of more than 9% of the predicted FEV1 as a hallmark for asthma confirmation
[1]. However, its results have limited reproducibility and accuracy. Since it depends
on several factors, such as the patient’s maximal effort and the bronchodilator
used [3].
Fluticasone/salmeterol combination increases FEV1, FEV1% of predicted, and FEV1/FVC
ratio than the conventional test using salbutamol inhaler, and it can be a potential
candidate for assessment of airway obstruction using reversibility test, especially
among the obese population.
Visser et al. (2015): “Reversibility of pulmonary function after inhaling
salbutamol in different doses and body postures in asthmatic children”

Bronchodilator reversibility Oscillometry
Jetmalani et al. (2021): “Normal limits for
oscillometric bronchodilator responses and
relationships with clinical factors”
We aimed to determine normal thresholds for
positive bronchodilator responses for oscillometry
in an Australian general population sample aged
≥40 years, to guide clinical interpretation. We also
examined relationships between bronchodilator
responses and respiratory symptoms, asthma
diagnosis, smoking and baseline lung function.
This study describes normative thresholds for
bronchodilator responses in oscillometry parameters,
including intra-breath parameters, as determined by
absolute, relative and Z-score changes. Positive
bronchodilator response by oscillometry correlated
with clinical factors and baseline function, which may
inform the clinical interpretation of oscillometry.
Or would there be even more passive (not requiring any effort / compliance from the patient) measure
for the bronchodilator reversibility? Sylvee-type of “simple” solution?
The bronchodilator-induced changes in a) resistance and b) reactance parameters of
the respiratory system, for each of the clinically defined groups. BDR: bronchodilator
response; ∆: post-bronchodilator minus pre-bronchodilator change; Rrs6: resistance of the
respiratory system measured at 6 Hz; insp: inspiratory; Xrs6: reactance of the respiratory system
measured at 6 Hz; EFLi: expiratory flow limitation index; Asym; asymptomatic: Symp:
symptomatic.

Asthma
Management
and Treatment

Individualize Treatment for subtype #1
https://preciseasthma.org/preciseweb/
Ray et al. (2022)

Individualize Treatment for subtype #2
Management of severe asthma in primary and secondary care – Jones et al. (2018)

mHealth Asthma Management
Tsang et al. (2021): “Application of Machine
Learning Algorithms for Asthma Management with
mHealth: A Clinical Review” Cited by 18
Asthma is a variable long-term condition. Currently,
there is no cure for asthma and the focus is,
therefore, on long-term management. Mobile
health (mHealth) is promising for chronic disease
management but to be able to realize its potential, it
needs to go beyond simply monitoring. mHealth
therefore needs to leverage machine learning to
provide tailored feedback with personalized
algorithms.
In the past five years, many studies have successfully
applied machine learning to asthma mHealth data.
However, most have been developed on small
datasets with internal validation at best. Small
sample sizes and lack of external validation limit
the generalizability of these studies. Future research
should collect data that are more representative of
the wider asthma population and focus on validating
the derived algorithms and technologies in a real-
world setting.

Pharma and
biomarkers in
clinical trials

Pharma Market and Biomarkers?
The AI-based Lung8 imaging software picked up differences in response between
patients who had an experimental drug compared with those who took a
placebo – key information that traditional analysis had failed to find, the
company noted.
Qureight chief executive officer Dr Muhunthan Thillai, a lung specialist at Royal
Papworth Hospital in Cambridge, said: "Existing clinical trials involving idiopathic
pulmonary fibrosis (IPF) can cost more than $150m per drug study and take at least 12
months to look for breathing changes in patients. Our novel platform technology
could halve the duration of trials and therefore significantly reduce the costs
involved. This will be invaluable for our pharmaceutical partners as they plan
their next set of drug trials for this complex disease.”
https://cordis.europa.eu/project/id/115010: The inability of pre-
clinical studies to predict clinical efficacy is a major bottleneck in drug
development. In severe asthma this bottleneck results from: a lack of
useful and validated biomarkers, underperforming pre-clinical models,
inadequate and incomplete sub-phenotyping, and insufficient disease
understanding.

Respiratory Biomarkers #1
https://www.appliedclinicaltrialsonline.com/view/breathing-new-life-into-respiratory-cli
nical-trials-with-digital-biomarkers
Respiratory conditions are some of the most pressing in terms of new
therapies, as the third leading cause of death worldwide. Clinical trials
for respiratory conditionsarealmost twiceas expensive as the average
clinical trial at an average of $91 million versus $48 million estimated
cost of trials per drug by therapeutic area. Digital biomarkers may be
one avenue to accelerate research and potentially bring study costs
down.
Digital biomarkers are measurements collected through a digital device,
such as portable, digital spirometers and smartwatches that can
track real-time changes in a person's health or behavior.
While often it’s necessary to develop entirely new digital biomarkers
when moving to remote data collection, that’s not usually the case in
respiratory research, since these biomarkers are often digitized, at-
home versions of the current gold standard. For example, a patient
will use a Bluetooth-enabled spirometer at home, rather than an analog
spirometer handed to them by a nurse in a clinic.
For respiratory studies, digital biomarkers offer a non-invasive way to monitor
respiratory function and provide objective data on symptoms and disease
progression in clinical trials. The tests can be conducted at home, in the
clinic, or used during at-home visits, eliminating travel-related patient
burdens. With more data points captured in a shorter period, a respiratory
clinical trial could potentially be completed faster without sacrificing quality.
With digital biomarkers, a patient can also undergo repeated testing.
By capturing a high volume of data, researchers can control for factors that
drive variation in spirometry measures, to potentially increase a study’s
statistical power and decreasing the number of patients needed to
demonstrate treatment effect. One example is the LEARN study, which is a
single arm interventional trial comparing the detection of treatment
effects by means of at-home mobile spirometry using an ultrasonic
spirometer and a smartphone, compared to in-clinic spirometry in
participants with moderate asthma.
For instance, repeat measurements from mobile spirometry have been
able to uncover information about diurnal variation in respiratory conditions
—the way symptoms change throughout the day and night. Digital
biomarkers used in remote assessments can also answer
the need for less subjective measures in respiratory trials. While variability
of at-home versus in-clinic measures has been a problem in the past,
studies have shown that at-home results can match in-clinic efforts.

Scott (2023): “Smartwatches in Respiratory Care: Ready for
Prime Time or Stick to Telling Time?”
With acknowledging the limitations of noninvasive monitoring,
researchers are continuously working to validate noninvasive
monitoring use in clinical settings. By understanding both the
benefits and limitations of evolving technology, clinicians can
effectively incorporate it into their practices.
However, keeping up with technologic advancements can be challenging,
even for those who consider themselves tech savvy. A relatively new
type of noninvasive technology that clinicians now must contend
with is wearables (eg, Apple Watch and Fitbit).
Liaqat et al. (2019): “WearBreathing: Real World Respiratory Rate
Monitoring Using Smartwatches”
We develop WearBreathing, a novel system for respiratory rate
monitoring. WearBreathing consists of a machine learning based
filter that detects and rejects sensor data that are not suitable for
respiratory rate extraction and a convolutional neural network
model for extracting respiratory rate from accelerometer and
gyroscope data.
IoT in Digital Respiratory Healthcare
Each chapter reviews contemporary literature and considers
not ‘if’ but ‘how’ a digital respiratory future can provide
optimal care. The result is an authoritative, balanced guide to
developing digital respiratory health.
Digital technology, including artificial intelligence, can (and increasing will)
contribute to all aspects of the “assess–adjust–review” cycle of personalised
asthma (self)-management. Specific digital contributions to supported self-
management include improving adherence to routine medication, checking and
correcting inhaler technique, monitoring asthma status, predicting risk, providing
timely advice via interactive action plans, enabling remote communication,
avoiding triggers and changing lifestyle behaviours. Implementation of digital
healthcare is a priority for professionals and healthcare systems but raises the
challenges of enabling connected integrated systems and avoiding increasing
inequities, and will require policy decisions on infrastructure and funding.

Marco-Ariño et al. (2021, AstraZeneca):
“Complexity of design in early phase respiratory
clinical trials; evaluating impact of primary
endpoints and sponsor size”
Asthma and COPD represent most of the
clinical trials in the respiratory area. The
Primary Endpoint (PE) defines how trials are
conducted. We hypothesised that small and
mid-sized pharmaceutical companies may be
innovative in the selection of their trial
endpoints, to be time- and cost-effective.
In conclusion, this exploratory analysis
indicates differences in study size, duration
and type of PE used by small/mid-sized and
large companies. For some types of
endpoints, differences in length and study size
were found. However, it wasn't possible to
attribute these differences between sponsors
solely to the choice of PE, pointing out to the
complexityofrunningclinicaltrials.

White et al. (2023, Boehringer Ingelheim): “Challenges for
Clinical Drug Development in Pulmonary Fibrosis” Cited by 25
The preclinical and clinical development of novel drug
candidates is hampered by profound challenges such as a
lack of sensitive clinical outcomes or suitable biomarkers that
would provide an early indication of patient benefit.
Liu et al. (2021): “Biomarkers for respiratory diseases: Present
applications and future discoveries”
Omics-based discovery strategies facilitate the identification of next-
generation candidate biomarkers. The identification and integration of
multiple omics biomarkers (genomics, epigenomics, proteomics,
metabolomics and radiomics) can answer many unsolved questions about
respiratory diseases. The ultimate realization of the translation of novel
biomarkers into clinical practice is a major challenge because good biomarkers
must be sensitive, technically feasible, non-invasive, non-expensive and most
importantly, validated. The four stages of clinical trial design in the
development of biomarkers are also discussed. The concept of systems
biology not only focuses on novel molecular biomarkers but also integrates
clinical, laboratory and omics information as a whole. Finally, developing multiple
biomarker panels is a goal to improve the healthcare of respiratory diseases.
Bime et al. (2020): “The acute respiratory distress
syndrome biomarker pipeline: crippling gaps between
discovery and clinical utility”
A critical gap exists between the fast pace of biomarker discovery
and the successful translation to clinical use. This gap underscores
the fundamental biomarker conundrum across various acute
and chronic disorders: how does a biomarker address a specific
unmet need? Additionally, the gap highlights the need to shift the
paradigm from a focus on biomarker discovery to greater
translational impact and the need for a more streamlined drug
approval process.
Tiotiu et al. (2018): “Biomarkers in asthma: state of the art”
The implementation of precision medicine in the management of
asthma requires the identification of phenotype-specific markers
measurable in biological fluids. To become useful, these
biomarkers need to be quantifiable by reliable systems,
reproducible in the clinical setting, easy to obtain and cost-
effective. Using biomarkers to predict asthma outcomes and
therapeutic response to targeted therapies has a great clinical
significance, particularly in severeasthma.

Leung and Sin (2013): “Biomarkers in airway
diseases”
The inherent limitations of spirometry and clinical
history have prompted clinicians and scientists to
search for surrogate markers of airway diseases.
Although few biomarkers have been widely
accepted into the clinical armamentarium, the
authors explore three sources of biomarkers that
have shown promise as indicators of disease
severity and treatment response. In asthma, exhaled
nitric oxide measurements can predict steroid
responsiveness and sputum eosinophil counts have
been used to titrate anti-inflammatory therapies. In
chronic obstructive pulmonary disease, inflammatory
plasma biomarkers, such as fibrinogen, club cell
secretory protein-16 and surfactant protein D, can
denote greater severity and predict the risk of
exacerbations. While the multitude of disease
phenotypes in respiratory medicine make biomarker
development especially challenging, these three may
soon play key roles in the diagnosis and
management of airway diseases.
Papi et al. (2021): “Rate of Decline of FEV1 as a Biomarker of
Survival?”
COPD has been defined by the excess decline in lung
function induced by tobacco smoking, with FEV1
considered the gold standard biomarker of COPD
development and progression
Dobler (2019): “Biomarkers in respiratory diseases”
The December issue of Breathe focuses on biomarkers in
respiratory diseases [Turner et al. 2019; Oliver et al. 2019;
Creamer et al. 2019; Hamerlijnck et al. 2019]. Biomarkers are
measurable indicators of the presence, severity or type of
a disease. They can help us understand the cause,
phenotype, progression or regression, prognosis, or outcome
of treatment of a disease. Biomarkers hold the promise of
personalised medicine, which aims to tailor treatments to
individual patients based on their biomarker profile and, by
doing so, reduce the harms from ineffective treatments
and increase the benefits from effective treatments.

Bousquet et al. (2023): “Patient-centered digital
biomarkers for allergic respiratory diseases and asthma:
The ARIA-EAACI approach – ARIA-EAACI Task Force
Report”
Biomarkers for the diagnosis, treatment and follow-up of patients with rhinitis
and/or asthma are urgently needed. Although some biologic biomarkers exist
in specialist care for asthma, they cannot be largely used in primary care.
Allergic Rhinitis and its Impact on Asthma (ARIA) and EAACI (European
Academy of Allergy and Clinical Immunology) developed a Task Force aimed
at proposing patient-reported outcome measures (PROMs) as digital
biomarkers that can be easily used for different purposes in rhinitis and
asthma. It first defined control digital biomarkers that should make a bridge
between clinical practice, randomized controlled trials (RCTs),
observational real-life studies and allergen challenges. Using the MASK-air
app as a model, a daily electronic combined symptom-medication score for
allergic diseases (CSMS) or for asthma (e-DASTHMA), combined with a
monthly control questionnaire, was embedded in a strategy similar to the
diabetes approach for disease control. To mimic real-life, it secondly
proposed quality-of-life digital biomarkers including daily EQ-5D visual
analogue scales and the bi-weekly RhinAsthma Patient Perspective (RAAP).
The potential implications for the management of allergic respiratory diseases
were proposed.
Applicability of digital biomarkers in severe asthma using the
diabetes approach.

Feb 21, 2024
https://www.medtechpulse.com/article/insight/ai-to-fight-severe-
respiratory-disease
In December, we brought you a story about
voice recordings. Specifically, we discussed a tool that
uses ten-second voice clips to screen for Type 2
diabetes.
This week, we’re back to highlight the power of audio
recordings to help screen for disease. But this time,
we’re talking about recording a sound that’s perhaps
more recognizable as a symptom of disease:
coughing.
In the past few weeks, headlines from Kashmir to Kenya
have reported on smartphone-based, AI-enabled
voice recording apps that can help providers screen for
one of the most dangerous respiratory diseases out
there—tuberculosis (TB).
MIT Technology Review identified 30–40 cough analysis
startups that came up due to the pandemic. Some, like
AudibleHealthAI, began due to COVID in 2020 and are now
expanding to diseases like TB and influenza. Others, like
ResApp Health, were around for years before COVID, but focused
their business on COVID screening in 2020 and found a solid
market foothold.
Plus, innovators have realized that, even if these tools can’t
distinguish between the causes of a cough, cough identification
in itself is useful.
Laryngologist Yael Bensoussan told MIT Technology Review that
where this technology can be especially useful is in clinical
trials, where researchers often have to rely on how well patients’
can remember and describe their coughing.
“It’s really hard to track cough,” she said. “It’s really easy to
capture the frequency of cough from a tech perspective.”

Digital Biomarkers in general #1

Digital Biomarkers in general #2
Powell (Feb 2024): “Walk, talk, think, see and feel:
harnessing the power of digital biomarkers in
healthcare”
This article explores the potential of common digital biomarkers and
their associated characteristics of health, namely, physical function
(walk), speech and acoustics (talk), cognitive (think), visual (see) and
wellbeing or symptoms (feel).
There is considerable optimism, interest, and development, but there is
a need to move from ambition to impact and to provide ‘value’ for
patients. To achieve this, more attention must be paid to integrating
and fusing multimodal data sources. Monitoring one impairment is
unlikely to reveal optimal insights for the diagnosis or monitoring of
complex traits or conditions (e.g., Dementia). By integrating digital
biomarkers with complementary data sources, such as ‘omics’ data
(multi-omics), we can usher in a new era of precision medicine guided
by deep phenotyping and endotyping..
This fusion is likely to provide richer insights on disease and has the
potential to unearth unique phenotypes or signatures that would allow
comparison within groups or pathological cohorts. Here we suggest
the term digital biomarker ‘fingerprints’ (DBfx) to encapsulate the
integration and fusion of multimodal data.
A new equation for healthcare: towards digital
biomarker ‘fingerprints’ (DBfx)

Digital Health in general in Pharma
Kaspar et al. (2024, strategy&): “A practical, experience-
based guide | Decoding digital health”
As discussed in our previous article (Decoding digital health: A success story for pharma?)
, we live in an era
where healthcare systems worldwide are confronted with unprecedented challenges,
and finding sustainable and innovative solutions is imperative. Over the last
decade, healthcare expenditure has surged by a staggering 24% across the EU,
according to the European Commission's 2022 report. Furthermore, a significant 35.2%
of people in the EU are living with chronic health problems, underscoring the urgent
need for transformative approaches.
As pharmaceutical companies (PharmaCos) navigate this landscape, digital health
solutions (DHS) are emerging as transformative tools, offering the promise of
enhanced patient outcomes and streamlined healthcare delivery. DHS represents a
groundbreaking fusion of information and communications technologies within the
medicine and healthcare professions, and incorporates the use of digital tools such
as telemedicine, remote monitoring, health platforms, and apps, all aimed at
managing illnesses, mitigating health risks, and promoting wellness.
Design with the data strategy in mind Data from DHS can be useful along a
PharmaCo’s value chain in many ways, — powering clinical trials, real-world evidence,
market research, and pharmacovigilance. An early, data-savvy strategy can
shape and secure a digital health solution's value proposition. Prioritizing user-
friendly, compliant, and safe data collection is a key driver for acceptance and buy-in.
As illustrated by Qualifyze, a pioneer in centralized pharma audits with a business
model built around their data, this strategic approach ensures that the data
generated is not only valuable but also aligns seamlessly with the overall goals of the
digital health solution.
I. Empathy-driven design: User-centric Digital Health Solution prototyping
II. Ecosystem validation: Establishing market alignment

AI in Clinical Trials
Hutson (2024, Nature): “How AI is
being used to accelerate clinical
trials”
Over the previous 60 years, the number of drugs
approved in the United States per billion dollars
in R&D spending had halved every nine years. It
can now take more than a billion dollars in
funding and a decade of work to bring one new
medication to market. Half of that time and
money is spent on clinical trials, which are
growing larger and more complex. And only one
in seven drugs that enters phase I trials is
eventually approved.
Now scientists are starting to use AI to manage
clinical trials, including the tasks of writing
protocols, recruiting patients and analysing data.
Reforming clinical research is “a big topic of
interest in the industry”, says Lisa Moneymaker,
the chief technology officer and chief product
officer at Saama, a software company in
Campbell, California, that uses AI to help
organizations automate parts of clinical trials. “In
terms of applications,” she says, “it’s like a kid in a
candy store.”
The first step of the clinical-trials
process is trial design. What dosages
of drugs should be given? To how
many patients? What data should be
collected on them? The lab of
Jimeng Sun, a computer scientist at
the University of Illinois Urbana-
Champaign, developed an algorithm
called HINT (hierarchical interaction
network) that can predict whether a
trial will succeed, based on the drug
molecule, target disease and patient
eligibility criteria. They followed up with
a system called SPOT (sequential
predictive modelling of clinical trial
outcome) that additionally takes into
account when the trials in its training
data took place and weighs more
recent trials more heavily. Based on
the predicted outcome,
pharmaceutical companies might
decide to alter a trial design, or try a
different drug completely.
A company called Intelligent Medical Objects in
Rosemont, Illinois, has developed SEETrials, a method
for prompting OpenAI’s large language model GPT-4 to
extract safety and efficacy information from the
abstracts of clinical trials. This enables trial designers
to quickly see how other researchers have designed
trials and what the outcomes have been. The lab of
Michael Snyder, a geneticist at Stanford University in
California, developed a tool last year called CliniDigest
that simultaneously summarizes dozens of records
from ClinicalTrials.gov, the main US registry for
medical trials, adding references to the unified
summary.
The most time-consuming part of a clinical trial is
recruiting patients, taking up to one-third of the study
length. One in five trials don’t even recruit the required
number of people, and nearly all trials exceed the
expected recruitment timelines. Some researchers
would like to accelerate the process by relaxing some
of the eligibility criteria while maintaining safety. A
group at Stanford led by James Zou, a biomedical data
scientist, developed a system called Trial Pathfinder
that analyses a set of completed clinical trials and
assesses how adjusting the criteria for participation
— such as thresholds for blood pressure and
lymphocyte counts — affects hazard ratios, or rates of
negative incidents such as serious illness or death
among patients.

Digital Therapeutics (DTx)
Miao et al. (March 2024): “Characterisation of
digital therapeutic clinical trials: a systematic
review with natural language processing”
https://github.com/BMiao10/ClinicalTrials + https://ct-gov.streamlit.app/
Digital therapeutics (DTx) are a somewhat novel class of
US Food and Drug Administration-regulated software that
help patients prevent, manage, or treat disease. Here, we
use natural language processing (BERTopic, SciSpacy) to
characterise registered DTx clinical trials and provide
insights into the clinical development landscape for these
novel therapeutics. We identified 449 DTx clinical trials,
initiated or expected to be initiated between 2010 and
2030, from ClinicalTrials.gov.
Examples of approved DTx include the Propeller
platform, which uses smart devices and paired
consumer applications to improve medication
adherence and reduces hospital admissions in
patients with asthma and chronic obstructive
pulmonary disease (COPD) (Moore et al. 2021;
Merchant et al. 2018)
Interventional DTx clinical trials by medical specialty
These can obviously include novel digital biomarkers, but could might as well be without

Hospital-at-home beyond just asthma
Miao et al. (March 2024): “The hospital at home in the USA:
current status and future prospects”
The annual cost of hospital care services in the US has risen to
over $1 trillion despite relatively worse health outcomes
compared to similar nations. These trends accentuate a
growing need for innovative care delivery models that reduce
costs and improve outcomes. HaH—a program that
provides patients acute-level hospital care at home—has
made significant progress over the past two decades.
Technological advancements in remote patient monitoring,
wearable sensors, health information technology
infrastructure, and multimodal health data processing have
contributed to its rise across hospitals. More recently, the
COVID-19 pandemic brought HaH into the mainstream,
especially in the US, with reimbursement waivers that made
the model financially acceptable for hospitals and payors.
However, HaH continues to face serious challenges to gain
widespread adoption. In this review, we evaluate the peer-
reviewed evidence and discuss the promises, challenges, and
what it would take to tap into the future potential of HaH.

‘Clinical
multimodal ML’
You probably want to
combine the lung sounds
with other clinical measures
Especially in clinical trials where the
lung sound be just one of the few
biomarkers explored?

EHR for asthma management #1
Alharbi et al. (2021): “Predictive models for
personalized asthma attacks based on
patient’s biosignals and environmental
factors: a systematic review”
Many researchers developed asthma attacks
prediction models that used various asthma
biosignals and environmental factors. These predictive
models can help asthmatic patients predict asthma
attacks in advance, and thus preventive measures
can be taken. Fifteen different asthma attack
predictive models were selected for this review.
Asthma attack predictive models become more
significant when using both patient’s biosignal and
environmental factors. There is a lack of utilizing
advanced machine learning methods, like deep
learning techniques. Besides, there is a need to build
smart healthcare systems that provide patients with
decision-making systems to identify risk and visualize
high-risk regions.
There were four works [23, 25, 33, 36] that used real-time computing to predict
asthma attacks per individual. Different wireless sensors were employed to
acquire data from users and the environment. Hosseini et al. [23] developed a
model to divide the risk of having an asthma episode into three categories (low,
medium, and high risk). The biosignals data were recorded through a built-in
smartwatch wireless sensor, while the environmental data were acquired
from different meteorological wireless sensors. Data were analyzed in real-
time through the cloud platform.

Budiarto et al. (2023): “Machine Learning–
Based Asthma Attack Prediction Models From
Routinely Collected Electronic Health Records:
Systematic Scoping Review”
An early warning tool to predict attacks could
enhance asthma management and reduce the
likelihood of serious consequences. Electronic health
records (EHRs) providing access to historical data
about patients with asthma coupled with machine
learning (ML) provide an opportunity to develop such
a tool. Several studies have developed ML-based tools
to predict asthma attacks.
Our review indicates that this research field is still
underdeveloped, given the limited body of evidence,
heterogeneity of methods, lack of external validation,
and suboptimally reported models. We highlighted
several technical challenges (class imbalance,
external validation, model explanation, and adherence
to reporting guidelines to aid reproducibility) that
need to be addressed to make progress toward
clinical adoption.

Huang and Huang (2023): “Use of feature
importance statistics to accurately predict
asthma attacks using machine learning: A cross-
sectional cohort study of the US population”
A cross-sectional analysis was conducted using a
modern, nationally representative cohort, the National
Health and Nutrition Examination Surveys (NHANES 2017–
2020). All adult patients greater than 18 years of age (total
of 7,922 individuals) with information on asthma attacks
were included in the study.
The top five highest ranked features by gain, a measure
of the percentage contribution of the covariate to the
overall model prediction, were Octanoic Acid intake as a
Saturated Fatty Acid (SFA) (gm), Eosinophil percent,
BMXHIP–Hip Circumference (cm), BMXHT–standing height
(cm) and HS C-Reactive Protein (mg/L).
Machine Learning models can additionally offer feature
importance and additional statistics to help identify
associations with asthma attacks.

‘Precision asthma’ pal?
Epic has now developed four programs for third-party vendors
that will be available in the 'showroom' (including its Open.Epic
API): cornerstone partners, pals, partners and member services.
Cornerstone partners, which now include Microsoft and
InterSystems, are companies that have technology and services
that Epic uses significantly in its software, Faulkner said.
"Partners" are established leaders in specific areas, according to
Epic executives. Last week, Epic confirmed that clinical
documentation company Nuance and survey software company
Press Ganey are the first companies to be in the "partners"
program.
The vendor relationships with Epic are not exclusive, Faulker
noted, as the vendors can work with other EHR companies and
Epic can work with other developers in the same area.
The "Pals" program focuses on early-stage products with
promising technology. Gen-AI-powered medical note-taking
service Abridge and digital contact center company Talkdesk
were tapped to be the first "pals."
https://www.fiercehealthcare.com/health-tech/epic-unveils-new-app-showroom-
third-party-vendors
https://vendorservices.epic.com/showroom

Multimodal Home Monitoring
Tsang et al. (2023): “Home monitoring with connected mobile
devices for asthma attack prediction with machine learning”
https://datashare.ed.ac.uk/handle/10283/4761
Monitoring asthma is essential for self-management. However,
traditional monitoring methods require high levels of active
engagement, and some patients may find this tedious. Passive
monitoring with mobile-health devices, especially when
combined with machine-learning, provides an avenue to reduce
management burden. Data for developing machine-learning
algorithms are scarce, and gathering new data is expensive. A few
datasets, such as the Asthma Mobile Health Study, are publicly
available, but they only consist of self-reported diaries and lack
any objective and passively collected data. To fill this gap, we
carried out a 2-phase, 7-month AAMOS-00 observational study to
monitor asthma using three smart-monitoring devices
(smart-peak-flow-meter/smart-inhaler/smartwatch), and
daily symptom questionnaires. Combined with localised weather,
pollen, and air-quality reports (ambee, OpenWeather), we
collected a rich longitudinal dataset to explore the feasibility of
passive monitoring and asthma attack prediction. This valuable
anonymised dataset for phase-2 of the study (device monitoring)
has been made publicly available. Between June-2021 and June-
2022, in the midst of UK’s COVID-19 lockdowns, 22 participants
across the UK provided 2,054 unique patient-days of data.
AAMOS-00 system overview. Centred around the participant’s own smartphone,
three smart monitoring devices (smart peak flow meter, smartwatch, smart
inhaler) collected objective data and two API services provided information about
local environment (weather, air quality, pollen) based on the participant’s location.

Lung Health
‘Covariate
tweaks’

Lung Age #1
Liang et al. (2022): “Estimation of lung age via a spline
method and its application in chronic respiratory diseases”
The concept of “lung age” was first proposed in 1985 to make
spirometry data easier to understand and as a physiological
instrument to evaluate the lung function damage caused by
smoking4
. A lung age older than the chronological age is
considered an indication of the accelerated decline or
impairment of lung function, and the difference between lung
age and chronological age (i.e., ∆ lung age) is used to
estimate the severity of this functional impairment. For example,
if a 50-year-old man has a lung age of 60 years old, then his ∆
lung age is 10 years, indicating there may be an impairment of
his lung function.
A randomized controlled trial showed that telling smokers
their lung age can improve the success rate of smoking
cessation5
. Lung age is also used as a clinical indicator in
studies regarding physiological changes in individuals with
morbid obesity6
and treatment efficacy in asthmatic patients7
.
However, it has not yet been widely used or fully exploited due
to the doubt regarding the current lung age estimation
methods8
.
Considering the potential application value of lung age in the
management of chronic respiratory diseases, this study aimed to
develop new lung age estimation equations and hypothesized
that nonlinear regression was a more appropriate method to
build the equations. Furthermore, the lung age of patients with
chronic obstructive pulmonary disease (COPD) and asthma was
estimated to explore the clinical application of lung age in
chronic respiratory diseases.
Comparisons of ∆ lung age between matched healthy subjects and patients
with COPD and asthma.

Lung Age #2 Cellular Senescence
Rongjun Wan et al. (2023): “Cellular senescence in asthma: from
pathogenesis to therapeutic challenges” Cited by 9
Asthma is a heterogeneous chronic respiratory disease that impacts nearly 10% of the
population worldwide. While cellular senescence is a normal physiological process, the
accumulation of senescent cells is considered a trigger that transforms physiology into
the pathophysiology of a tissue/organ. Recent advances have suggested the significance
of cellular senescence in asthma.
With this review, we focus on the literature regarding the physiology and pathophysiology of
cellular senescence and cellular stress responses that link the triggers of asthma to
cellular senescence, including telomere shortening, DNA damage, oncogene activation,
oxidative-related senescence, and senescence-associated secretory phenotype (SASP). The
association of cellular senescence to asthma phenotypes, airway inflammation and
remodeling, was also reviewed. Importantly, several approaches targeting cellular
senescence, such as senolytics and senomorphics, have emerged as promising strategies
for asthma treatment. Therefore, cellular senescence might represent a mechanism in
asthma, and the senescence-related molecules and pathways could be targeted for
therapeutic benefit.
Recent advances suggest cellular senescence as a mechanism in airway inflammation. While the
mechanisms remain unclear, developmental and replicative senescence have been suggested to play
a vital role in airway inflammation.12
Stress-induced senescence in response to stressors such as
pollution, smoking, or allergens causes the secretion of SASP, thereby leading to an increased airway
inflammation.40
Furthermore, different cell types, including structural and immune cells, may contribute
to senescence and airway inflammation in response to various stressors.
DAMPs (Damage-associated molecular patterns)

‘Biological Age’ #1
https://doi.org/10.1038/d41586-019-02638-w
https://doi.org/10.1093/gerona/glad174
https://doi.org/10.1038/s42003-023-05456-z
https://doi.org/10.3389/fgene.2020.630186
Environmental and lifestyle factors (stress, exercise, obesity, smoking) can age the patient more/less compared to their
chronological time. Something to be aware at least if you use age as a clinical covariate.
https://doi.org/10.1183/13993003.01806-2023
Telomere length is shown to be fundamental to a
person's outcome and their response to
treatment, albeit an adverse response to
immunosuppression, the authors consider
telomere length as a variable that modulates
outcome in an individual with a correctly
diagnosed ILD, rather than a parameter which
might instead inform the diagnosis.
https://doi.org/10.14336/ad.2020.1202

‘Biological Age’ #2
Shruthi Hamsanathan et al. (2024): “A molecular index for
biological age identified from the metabolome and
senescence-associated secretome in humans”
Unlike chronological age, biological age is a strong indicator of
health of an individual. However, the molecular fingerprint
associated with biological age is ill-defined. To define a high-
resolution signature of biological age, we analyzed metabolome,
circulating senescence-associated secretome (SASP)/inflammation
markers and the interaction between them, from a cohort of healthy
and rapid agers. The balance between two fatty acid oxidation
mechanisms, -oxidation and -oxidation, associated with the
β ω
extent of functional aging. Furthermore, a panel of 25 metabolites,
Healthy Aging Metabolic (HAM) index, predicted healthy agers
regardless of gender and race. HAM index was also validated in an
independent cohort. Causal inference with machine learning implied
three metabolites, -cryptoxanthin, prolylhydroxyproline, and
β
eicosenoylcarnitine as putative drivers of biological aging.
Multiple SASP markers were also elevated in rapid agers. Together,
our findings reveal that a network of metabolic pathways underlie
biological aging, and the HAM index could serve as a predictor of
phenotypic aging in humans.

‘Whatever Age’: Eye Age
Wolf et al. (2023): “Liquid-biopsy proteomics combined with AI
identifies cellular drivers of eye aging and disease in vivo”
Single-cell analysis in living humans is essential for understanding
disease mechanisms, but it is impractical in non-regenerative
organs, such as the eye and brain, because tissue biopsies would
cause serious damage. We resolve this problem by integrating
proteomics of liquid biopsies with single-cell transcriptomics from
all known ocular cell types to trace the cellular origin of 5,953
proteins detected in the aqueous humor. We identified hundreds of
cell-specific protein markers, including for individual retinal cell types.
Surprisingly, our results reveal that retinal degeneration occurs in
Parkinson’s disease, and the cells driving diabetic retinopathy switch
with disease stage. Finally, we developed artificial intelligence (AI)
models to assess individual cellular aging and found that many eye
diseases not associated with chronological age undergo
accelerated molecular aging of disease-specific cell types. Our
approach, which can be applied to other organ systems, has the
potential to transform molecular diagnostics and prognostics while
uncovering new cellular disease and aging mechanisms.

‘Whatever Age’
Organ clocks are an important advance as we move
toward the potential of modulating human biological
aging wsj.com by @AlexLJanin This article is about
@wysscoray's outstanding work, published @Nature,
that we recently discussed erictopol.substack.com
Jarod Rutledge, Ph.D.
https://twitter.com/JarodRutledge1/status/1733272252513398957

Sex/gender and asthma management
Boulet et al. (2022): “Addressing sex
and gender to improve asthma
management” Cited by 8
Sex (whether one is ‘male’ or ‘female’, based on
biological characteristics) and gender (defined by
socially constructed roles and behaviors) influence
asthma diagnosis and management. For example,
women generally report more severe asthma
symptoms than men; men and women are
exposed to different asthma-causing triggers; men
tend to be more physically active than women.
Furthermore, implicit, often unintended gender bias
by healthcare professionals (HCPs) is widespread,
and may result in delayed asthma diagnosis,
which can be greater in women than men. The sex
and gender of the HCP can also impact asthma
management. Pregnancy, menstruation, and
menopause can all affect asthma in several ways
and may be associated with poor asthma control.
This review provides guidance for considering sex-
and gender-associated impacts on asthma
diagnosis and management and offers possible
approaches to support HCPs in providing
personalized asthma care for all patients,
regardless of their sex or gender.

Asthma Phenotypes #1
Wenzel (2012): “Asthma phenotypes: the
evolution from clinical to molecular approaches”
Cited by 2894
Although asthma has been considered as a
single disease for years, recent studies have
increasingly focused on its heterogeneity. The
characterization of this heterogeneity has
promoted the concept that asthma consists of
multiple phenotypes or consistent groupings of
characteristics. Asthma phenotypes were initially
focused on combinations of clinical
characteristics, but they are now evolving to link
biology to phenotype, often through a statistically
based process. Ongoing studies of large-scale,
molecularly and genetically focused and
extensively clinically characterized cohorts of
asthma should enhance our ability to molecularly
understand these phenotypes and lead to more
targeted and personalized approaches to
asthma therapy.

Howard et al. (2015): “Distinguishing Asthma Phenotypes
Using Machine Learning Approaches” Cited by 127
Asthma is not a single disease, but an umbrella term for a
number of distinct diseases, each of which are caused by a
distinct underlying pathophysiological mechanism. These
discrete disease entities are often labelled as 'asthma
endotypes'. The discovery of different asthma subtypes has
moved from subjective approaches in which putative
phenotypes are assigned by experts to data-driven ones which
incorporate machine learning. This review focuses on the
methodological developments of one such machine learning
technique-latent class analysis-and how it has contributed to
distinguishing asthma and wheezing subtypes in childhood. It
also gives a clinical perspective, presenting the findings of
studies from the past 5 years that used this approach. The
identification of true asthma endotypes may be a crucial step
towards understanding their distinct pathophysiological
mechanisms, which could ultimately lead to more precise
prevention strategies, identification of novel therapeutic targets
and the development of effective personalized therapies.
Further research will be necessary in order to replicate different asthma, wheeze and atopy
subtypes across independent cohorts, to assess their stability over time and to confirm the
existence of distinct pathophysiological mechanisms underpinning each sub-type. Since
unique pathophysiological mechanisms for the subtypes of wheezing illness and atopy
identified so far using machine learning approaches have not as yet been elucidated, these
cannot be considered as ‘true endotypes’ but are mostly hypothetical constructs to facilitate
further research in this area. The identification of underlying biology and real endotypes may
have major implications for effective and precise asthma prevention, treatment and
management strategies, as it is anticipated that the different groups may respond differently
to the treatments currently offered.

Zhan et al. (2023): “Identification of cough-variant asthma
phenotypes based on clinical and pathophysiologic data”
Cough-variant asthma (CVA) is a subtype of asthma that usually
presents solely with cough without any other symptoms such as
dyspnea or wheezing [Corrao et al. 1979]. In cough-predominant
asthma cough is the most predominant symptom but other symptoms
are also present such as dyspnea and/or wheeze [Niimi 2008]. CVA may
respond differently to antiasthmatic treatment. There are limited data
on the heterogeneity of CVA. We aimed to classify patients with CVA
using cluster analysis based on clinicophysiologic parameters and to
unveil the underlying molecular pathways of these phenotypes with
transcriptomic data of sputum cells.
Cluster 1 was characterized by female predominance, late onset,
normal lung function, and a low proportion of complete resolution of
cough (60.8%) after antiasthmatic treatment. Patients in cluster 2
presented with young, nocturnal cough, atopy, high type 2
inflammation, and a high proportion of complete resolution of cough
(73.3%) with a highly upregulated coexpression gene network that
related to type 2 immunity. Patients in cluster 3 had high body mass
index, long disease duration, family history of asthma, low lung
function, and low proportion of complete resolution of cough (54.1%).

from Phenotypes to Endotypes #1
Overview on data used to
define different phenotypes
and endotypes.
Heterogeneous asthmatic
children are depicted on the
top of the figure. Data type
level specifies which kind of
data must at least be present
to identify specific groups
resulting in clinical
phenotypes, inflammatory
endotypes, and molecular
endotypes
Foppiano and Schaub (2023)
“Childhood asthma
phenotypes and endotypes:
a glance into the mosaic”

Bagnasco et al. (2019): “Evolving phenotypes to
endotypes: is precision medicine achievable in
asthma?”
The development of biologic molecules led to a drastic change in
the therapeutic approach to asthma. With the prospect of acting
on different pathophysiological mechanisms of the disease, the
idea of precision medicine was developed, in which a single
molecule is able to modify a specific triggering mechanism. Thus,
it seemed limiting to stop at the distinction of patients phenotypes
and the concept of endotypes became more relevant in the
therapeutic approach.
The possible overlap of patients in different phenotypes
requires a more precise classification, which considers
endotypization. With the development of biological drugs able to
modify and modulate some pathophysiological mechanisms of
the disease, the theoretical concept of endotyping becomes
practical, allowing the clinician to choose the specific mechanism
to ‘attack’ in order to control the disease.
An endotype is a subtype of a health condition, which is defined by a distinct functional or pathobiological mechanism. This is distinct from a
phenotype, which is any observable characteristic or trait of a disease, such as development, biochemical or physiological properties without any
implication of a mechanism.

Tang et al. (2020): “Systems biology and big data in
asthma and allergy: recent discoveries and emerging
challenges”
Systems biology is a relatively recent development that addresses the
growing complexity of biomedical research questions. Variation in inherited
susceptibility and exposures causes heterogeneity in manifestations of
asthma and other allergic diseases. Machine learning approaches are
being used to explore this heterogeneity, and to probe the
pathophysiological patterns or “endotypes” that correlate with
subphenotypes of asthma and allergy. Mathematical models are being
built based on genomic, transcriptomic and proteomic data to predict or
discriminate disease phenotypes, and to describe the biomolecular
networks behind asthma.
Worldwide, there has been a push by many groups to implement systems medicine, charting a path from wet
lab to dry lab to bedside. Large consortia, such as MeDALL (Mechanisms of the Development of Allergy) in Europe
and STELAR (Study Team for Asthma Life Research) from the UK, have been established specifically to record and
integrate multi-omic data related to allergy and asthma, and conduct well-powered systems-based analyses.
Other smaller groups are also involved in similar research via frequent cross-collaborations: these include the
Childhood Asthma Study (Australia), U-BIOPRED (Unbiased Biomarkers in Prediction of Respiratory Disease
Outcomes; European), Childhood Origins of Asthma (USA), Copenhagen Prospective Study on Asthma in Childhood
(COPSAC; Denmark), Manchester Asthma and Allergy Study (UK), Severe Asthma Research Program (USA) and
others. In the modern age of systems biology, collaboration and data sharing is virtually mandatory when it
comes to uncovering complex associations such as gene–environment interactions.

Sadegh et al. (2023): “Lacking mechanistic disease
definitions and corresponding association data
hamper progress in network medicine and beyond”
Cited by 3
A long-term objective of network medicine is to replace our
current, mainly phenotype-based disease definitions by
subtypes of health conditions corresponding to distinct
pathomechanisms. For this, molecular and health data are
modeled as networks and are mined for pathomechanisms.
However, many such studies rely on large-scale disease
association data where diseases are annotated using the
very phenotype-based disease definitions the network
medicine field aims to overcome. This raises the question to
which extent the biases mechanistically inadequate disease
annotations introduce in disease association data distort the
results of studies which use such data for pathomechanism
mining. We address this question using global- and local-
scale analyses of networks constructed from disease
association data of various types. Our results indicate that
large-scale disease association data should be used with
care for pathomechanism mining and that analyses of such
data should be accompanied by close-up analyses of
molecular data for well-characterized patient cohorts.

Getting Asthma Endotypes #1
Ray et al. (2022): “Determining asthma endotypes and
outcomes: Complementing existing clinical practice with
modern machine learning” Cited by 12
There is unprecedented opportunity to use machine learning to
integrate high-dimensional molecular data with clinical
characteristics to accurately diagnose and manage disease.
Asthma is a complex and heterogeneous disease and cannot
be solely explained by an aberrant type 2 (T2) immune response.
Available and emerging multi-omics datasets of asthma show
dysregulation of different biological pathways including those
linked to T2 mechanisms. While T2-directed biologics have been
life changing for many patients, they have not proven effective
for many others despite similar biomarker profiles. Thus, there
is a great need to close this gap to understand asthma
heterogeneity, which can be achieved by harnessing and
integrating the rich multi-omics asthma datasets and the
corresponding clinical data. This article presents a compendium
of machine learning approaches that can be utilized to bridge
the gap between predictive biomarkers and actual causal
signatures that are validated in clinical trials to ultimately
establish true asthma endotypes.

Getting Asthma Endotypes #2
Agache et al. (2023): “Multidimensional endotyping using
nasal proteomics predicts molecular phenotypes in the
asthmatic airways” Cited by 12
Unsupervised clustering of biomarkers derived from noninvasive
samples such as nasal fluid is less evaluated as a tool for
describing asthma endotypes. We sought to evaluate whether
protein expression in nasal fluid would identify distinct clusters
of patients with asthma with specific lower airway molecular
phenotypes.
The nasal protein machine learning model identified 2 severe
asthma endotypes, which were replicated in U-BIOPRED nasal
transcriptomics. Cluster 1 patients had significant airway
obstruction, small airways disease, air trapping, decreased
diffusing capacity, and increased oxidative stress, although only
4 of 18 were current smokers.
Protein or gene analysis may indicate molecular phenotypes
within the asthmatic lower airway and provide a simple,
noninvasive test for non–type 2 immune response asthma
that is currently unavailable.
Unsupervised
clustering (UMAP +
GMM + Shapley
ranking, Shap values)

Profiling for Biologics
van der Burg and Tufvesson (2023): “Is asthma's
heterogeneity too vast to use traditional phenotyping
for modern biologic therapies?”
Clinical trials of novel biologics and cluster analysis in asthma
still heavily rely on clinically-defined phenotypes. Biologics are
unique in that they target a specific antibody, molecule, or cell
involved in asthma. Because of this, they are known as
“precision” or “personalized” therapy. Bioprofiling patients with
asthma could provide more targetable groups for modern
biologics.
Quality-of-life questionnaire scores consistently show how
patients improve when they perceive to have less asthma
symptoms, sleep problems, anxiety and depression as well as
experience the regained capability to actively participate in
daily life. However, such reports of asthma do not include precise
measurements that reflect the underlying mechanisms of the
disease nor the severity of the airway inflammation in individual
patients. Meanwhile, clinical and real-life evidence show that a
substantial number of asthma patients appear to have relative
corticosteroid resistance and thus remain uncontrolled when
treated with the one-size-fits-all approach.
Unfortunately, to date, clinical trials investigating novel biologics in asthma heavily
rely on clinically defined phenotypes as target populations, often without a clear link
to the underlying biological profile (at treatment initiation nor over time) within these
target populations[[19],[20]]
. In the majority of such trials, T2-high severe asthma has
become the primary target population because of the availability of some
biomarkers that can be used to determine this phenotype[[16],[19]]
. Vice-versa, there is a
strong bias to developing T2 targeted biologics[[21]]
, leaving the T2-low and non-T2
mechanisms largely undefined and untargeted[[20]]
. Even within the T2-high asthma
phenotype landscape, the response to existing biologics is unpredictable and can be
associated with a substantial proportion of non-responders or partial responders[[22],
[23]]
.
There is a lot to be investigated in the context of biologically-based clustering of
asthma patients. Changes in the clinico-bioprofile must also be assessed, over time
with and without the intervention of biologics. There are also no clear indications that
current clustering techniques are stable within a patient over time nor did any multi-
center studies assess site-to-site specific effects on clusters. Future studies should
consider generating this data as it will become highly useful when we first begin to
form unbiased cluster groups of patients with asthma for selection of biologics both in
the context of clinical trials as well as in clinical practice.
By using longitudinal clinico-biological data from large asthma registries and
defining biologically-based clusters of asthma patients, we believe the development
and application of biologics will become more precise to asthma patients as
individuals.

Precision Therapy
When one have figured the
endotype of your patient, one
can tailor the treatment for
the patient

Precision Medicine in Asthma Therapy #1
Chung and Adcock (2019): “Precision medicine for the discovery of
treatable mechanisms in severe asthma” Cited by 93
The pharmacological approach to initiating treatment has, until recently, been
based on disease control and severity. The introduction of antibody therapies
targeting the Type 2 inflammation pathway for patients with severe asthma has
resulted in the recognition of an allergic and an eosinophilic phenotype, which
are not mutually exclusive. Concomitantly, molecular phenotyping based on a
transcriptomic analysis of bronchial epithelial and sputum cells has identified a
Type 2 high inflammation cluster characterized by eosinophilia and recurrent
exacerbations, as well as Type 2 low clusters linked with IL-6 trans-signalling,
interferon pathways, inflammasome activation and mitochondrial oxidative
phosphorylation pathways.
Systems biology approaches are establishing the links between these pathways
or mechanisms, and clinical and physiologic features. Validation of these
pathways contributes to defining endotypes and treatable mechanisms. Precision
medicine approaches are necessary to link treatable mechanisms with treatable
traits and biomarkers derived from clinical, physiologic and inflammatory features
of clinical phenotypes. The deep molecular phenotyping of airway samples along
with noninvasive biomarkers linked to bioinformatic and machine learning
techniques will enable the rapid detection of molecular mechanisms that
transgresses beyond the concept of treatable traits.

Precision Medicine in Asthma Therapy #2
Cazzola et al. (2023): “Revisiting asthma pharmacotherapy: where do we
stand and where do we want to go?”
Several current guidelines/strategies outline a treatment approach to asthma,
which primarily consider the goals of improving lung function and quality of life
and reducing symptoms and exacerbations. They suggest a strategy of stepping
up or down treatment, depending on the patient's overall current asthma
symptom control and future risk of exacerbation. While this stepwise approach is
undeniably practical for daily practice, it does not always address the
underlying mechanisms of this heterogeneous disease.
In the last decade, there have been attempts to improve the treatment of severe
asthma, such as the addition of a long-acting antimuscarinic agent to the
traditional inhaled corticosteroid/long-acting 2-agonist treatment and the
β
introduction of therapies targeting key cytokines. However, despite such strategies
several unmet needs in this population remain, motivating research to identify
novel targets and develop improved therapeutic and/or preventative asthma
treatments. Pending the availability of such therapies, it is essential to re-
evaluate the current conventional “one-size-fits-all” approach to a more
precise asthma management. Although challenging, identifying “treatable
traits” that contribute to respiratory symptoms in individual patients with asthma
may allow a more pragmatic approach to establish more personalised
therapeutic goals.

Chronotherapy for Chronic Pulmonary Diseases
Giri et al. (2022): “Circadian clock-based therapeutics in chronic
pulmonary diseases” Cited by 8
Circadian molecular clock disruption occurs due to
dysregulated immune–inflammatory response in the lung
which further contributes to the pathobiology of chronic
pulmonary diseases.
Recent evidence from preclinical models strongly suggests
that selective manipulation of core circadian clock
molecules (chronopharmacology) may contribute to the
development of novel clock-based therapeutics against
chronic lung diseases.
Preclinical studies support that pharmacological
manipulation of the circadian clock is a conceivable
approach for the development of novel clock-based
therapeutics. Despite recent advances, no effort has been
undertaken to integrate novel findings for the treatment
and management of chronic lung diseases. We, therefore,
recognize the need to discuss the candidate clock genes
that can be potentially targeted for therapeutic
intervention. Here, we aim to create the first roadmap that
will advance the development of circadian- clock-based
therapeutics that may provide better outcomes in
treating chronic pulmonary diseases.

Treatable Traits #1a
Gibson et al. (2023): “Treatable traits, combination inhaler
therapy and the future of asthma management” Cited by 1
Currently, inhaled corticosteroids (ICS), with or without long-acting beta
agonists (LABA), are the cornerstone of asthma management, and recently
international guidelines recognized the importance of combination inhaler
therapy (ICS/LABA) even in mild asthma. In future, ultra-long-acting
personalized medications and smart inhalers will complement combination
inhaler therapy in order to effectively addresses issues such as adherence,
inhaler technique and polypharmacy (both of drugs and devices). Asthma is
now acknowledged as a multifaceted cluster of disorders and the treatment
model has evolved from one-size-fits-all to precision medicine approaches
such as treatable traits (McDonald et al. 2019, Thomas 2022
; TTs, defined as measurable
and treatable clinically important factors) which encourages the quality use
of medications and identification and management of all underlying
behavioural and biological treatable risk factors.
TT requires research and validation in a clinical context and the
implementation strategies and efficacy in various settings
(primary/secondary/tertiary care, low-middle income countries) and
populations (mild/moderate/severe asthma) are currently evolving.
Combination inhaler therapy and the TTs approach are complementary
treatment approaches. This review examines the current status of personalized
medicine and combination inhaler therapy, and describes futuristic views for
these two strategies.
What is a treatable trait? An identifiable feature that can be assessed
and targeted by treatment to improve a clinical outcome is called a
treatable trait.20
Three characteristics are present in each treatable trait
10, 21, 22
: (1) clinical significance (i.e., the trait is associated with a clinical
outcome, for example, asthma exacerbation, quality of life [QoL] or
asthma control); (2) detectable (i.e., measurable via specific and
validated ‘trait identification markers [TIM]’, e.g., the trait of T2
inflammation is assessed by the TIM of circulating eosinophils); and,
(3) treatable (an effective treatment is available).
Eosinophilic/T2 airway inflammation is an excellent example of a
treatable trait in asthma. This endotype is mediated by specific cytokines
such as interleukin (IL)-5.23
The levels of eosinophilic inflammation are
directly proportional to exacerbation risk indicating its clinical relevance.
Moreover, it can be detected via blood eosinophil count (a TIM) and can
be treated via corticosteroids (inhaled or oral) or T2-targeted biologics.
Hence, eosinophilic airway inflammation meets all the criteria for a
treatable trait (i.e., relevant, detectable and treatable).
Treatable traits approach has already been described for use in tertiary
care settings. The future development of treatable traits will identify how
to adapt treatable traits to different settings, such as primary care.
Lung sounds / other biomarkers the (2) of treatable traits (“detectable”)

Treatable Traits #1b
Gibson et al. (2023): “Treatable traits, combination
inhaler therapy and the future of asthma
management”

Treatable Traits #2
McDonald and Holland (2024, Australia):
“Treatable traits models of care”
Treatable traits is a personalized approach to the
management of respiratory disease. The approach
involves a multidimensional assessment to
understand the traits present in individual patients.
Traits are phenotypic and endotypic characteristics
that can be identified, are clinically relevant and can
be successfully treated by therapy to improve clinical
outcomes. Identification of traits is followed by
individualized and targeted treatment to those traits.
There are several key aspects to treatable traits
models of care that should be considered in the
delivery of care. These include person centredness,
consideration of patients' values, needs and
preferences, health literacy and engagement. Target one trait to treat multiple traits—this is an example of how effectively targeting a
super trait, in this case adherence, has a multiplicative effect by reducing the impact of
multiple other traits and improving overall goals of treatment

Treatable Traits #3
Pérez de Llano et al. (2019): “Phenotype-Guided Asthma Therapy: An Alternative Approach to Guidelines”
A clinical example of therapeutic goals and treatable traits.

Variability from poor adherence and poor inhaler skills #1
https://doi.org/10.1002/14651858.CD013030.pub2 - Cited by 44
https://doi.org/10.4187/respcare.06740

Variability from poor adherence and poor inhaler skills #2
van de Hei (2023): “Anticipated barriers and
facilitators for implementing smart inhalers in
asthma medication adherence management”
Smart inhalers are electronic monitoring devices
which are promising in increasing medication
adherence and maintaining asthma control. A
multi-stakeholder capacity and needs assessment
is recommended prior to implementation in
healthcare systems. This study aimed to explore
perceptions of stakeholders and to identify
anticipated facilitators and barriers associated with
the implementation of smart digital inhalers in the
Dutch healthcare system. Data were collected
through focus group discussions with female
patients with asthma (n = 9) and healthcare
professionals (n = 7) and through individual semi-
structured interviews with policy makers (n = 4) and
smart inhaler developers (n = 4). Data were
analysed using the Framework method. Five themes
were identified: (i) perceived benefits, (ii) usability,
(iii) feasibility, (iv) payment and reimbursement, and
(v) data safety and ownership. In total, 14 barriers
and 32 facilitators were found among all
stakeholders. The results of this study could
contribute to the design of a tailored
implementation strategy for smart inhalers in daily
practice.
Key factors that should be addressed prior to large-
scale implementation include the lack of evidence of
effectiveness and target groups, usability issues,
workflow changes in healthcare organisations due
to implementation, compatibility issues on different
levels, the lack of reimbursement agreements and
clarity regarding data access and security. The results
of this study could contribute to the design of an
implementation strategy and may help developers in
improving the design of the current devices, which
together could increase the chance of successful
implementation.

Cough and cough hypersensitivity as treatable traits of asthma
Cazzola et al. (2023): “Revisiting asthma
pharmacotherapy: where do we stand and where do
we want to go?”
Cough is a common and troublesome symptom in people with
asthma and is often associated with poorer asthma control and
exacerbations. Apart from asthma, other causes or comorbidities
might underlie cough in asthma, such as rhinosinusitis and
bronchiectasis. Eosinophilic inflammation and bronchoconstriction
can lead to an acute episode of cough or worsen chronic cough.
Cough hypersensitivity with laryngeal paraesthesia, allotussia,
and hypertussia might underlie the cough of asthma through
augmented sensory nerve excitability of upper-airway vagal
sensory nerves. Cough associated with bronchoconstriction and
type 2 inflammation should respond to inhaled corticosteroids
and long-acting -adrenoceptor agonist therapy. For cough
β
hypersensitivity in adults, speech and language therapy and
neuromodulators (eg, gabapentin) could be considered. In
children, there is no consistent association of asthma with cough
sensitivity or between cough and asthma severity. Further
research is needed to realise the potential of cough as a
measure of asthma control, to understand the mechanisms of
cough in asthma, and to develop safe, effective treatments and a
precision-medicine approach to the management of cough in
asthma in children and adults.

Clinical Inertia in asthmaa
Fukuda et al. (2023): “Clinical inertia in asthma”
Despite advances in pharmaceutical
treatment in recent years, a relatively high
proportion of patients with asthma do not
have adequate asthma control, causing
chronic disability, poor quality of life, and
multiple emergency department visits and
hospitalizations. A multifaceted approach is
needed to overcome the problems with
managing asthma, and clinical inertia (CI) is
a crucial concept to assist with this approach.
It divides clinical inertia into three main
categories, which include healthcare
provider-related, patient-related, and
healthcare system-related CI. The strategies
to overcome these CI are complex, and the
M-GAP approach, which combines a
multidisciplinary approach, dissemination of
guidelines, utilization of applications, and
development and promotion of low-cost
prescriptions, will help clinicians.
The concept of CI was proposed by Phillips et al.17
in 2001 and has been applied
mainly to lifestyle-related diseases such as hypertension, dyslipidemia, and
diabetes. Phillips et al.17
defined CI as the failure of healthcare providers to initiate
or intensify appropriate treatment despite various therapeutic advances having
clarified treatment goals. In an observational study of approximately 1500 patients
with hypertension treated by 500 physicians, 30% of the participants had CI that
needed to be corrected18
. Even when telemonitoring triggered an intervention alert
that the average home blood pressure was elevated over 2 weeks, more than half
of the physicians reported that they did not intensify treatment because they
judged the blood pressure to be within an acceptable range19
.
A Spanish study of patients with type 2 diabetes found that CI was present in one-
fifth to one-quarter of patients and that the prevalence varied according to the
patient’s hemoglobin A1c level21
. Okonofua et al.22
defined this condition as
therapeutic inertia (TI) and used it to stratify patients. They found that in patients
with hypertension, a higher TI score was correlated with poorer blood pressure
control. In patients with lifestyle-related diseases, such as treatable
cardiovascular risk factors, these factors affect the clinical outcomes. CI is a
significant problem that needs to be overcome by patients, their families, and
healthcare providers for patients to maintain their quality of life and lead healthy
lives.
When “What gets measured gets managed” is not happening

Clinical Inertia in asthma b
Fukuda et al. (2023): “Clinical inertia in asthma”
CI is also present in asthma practice. In some
cases, asthma is not controlled, yet appropriate
therapeutic intervention or intensification is not
provided, or treatment that is no longer needed is
continued without reducing the dose or
discontinuation. One study found that 39% of
patients diagnosed with “definite asthma”
were not receiving any medication23
. Another
study found that 39% of patients with uncontrolled
asthma and 53% of patients with partially
controlled asthma were not using daily
medication for asthma control, equivalent to GINA
treatment step 115
. CI is not necessarily caused by
a single factor but is due to a complex interplay
of factors and needs to be addressed using a
multifaceted approach.
Factors affecting CI in asthma fall into three main
categories: those related to the healthcare
provider (contribution rate: 50%), those related
to the patient (contribution rate: 30%), and those
related to the healthcare system (contribution
rate: 20%), although some factors are shared
between categories24
(Fig. 1).
Fig. 1: Clinical inertia associated with asthma clinical practice. CIs include healthcare-related,
patient-related, and healthcare system-related. Each CI consists of multiple elements. CI clinical inertia.
CI applies not only to an intensification of treatment but also to follow-up of patients who are discontinuing treatment
or transitioning to a lower dose, including discontinuing medications that are no longer needed90
. Patients on multiple
medications are a prime example. One study found that more than half of older patients wanted to reduce the
number of medications that they were taking if their doctors deemed it possible91
. A comparison of patients on multiple
medications with those not on multiple medications showed that asthma control was predominantly worse in
patients on multiple medications33
. The same is true for asthma overdiagnosis. Aaron et al.92
reported that among
patients diagnosed with asthma during the past 5 years, approximately 30% of patients had no evidence of asthma
when reevaluated using symptom monitoring, spirometry, and peak flow. Similarly, in a study conducted in Sweden, 34%
of patients diagnosed with asthma were found not to have asthma when evaluated by an allergist based on
respiratory function tests and a methacholine challenge test93
. This suggests the importance of distinguishing
between pure CI and “appropriate inaction” in asthma care94
. Therefore, it is crucial for healthcare providers to
improve their practice skills by resolving CI using an M-GAP approach and for patients to improve their self-
management skills.

Reimbursement Who pays for this?
Since the new icometrix CPT codes (0865T/0866T) were
issued in the US, we’ve been bombarded with requests
from hospitals that like to start implementing the code.
The novel reimbursement pathway is a game-changer
for neuroradiology and neurology. Besides access to
clinically meaningful and quantified brain MRI
information (#icobrain) that impacts diagnostic and
therapeutic decision-making in individual patients, the
CPT code provides a sustainable business model for more
widespread adoption and access.
The story of these new CPT codes hinges on #innovation,
#collaboration, and placing #patientwelfare at the
forefront. The approval and implementation of these new
CPT codes isn't just a win for icometrix; it represents a
significant leap towards integrating cutting-edge
diagnostics and treatment monitoring into patient care
and transforming them from a privilege to a universal
standard of care.
March 8, 2024
https://www.linkedin.com/pulse/new-cpt-code-brain-mri-quantification-wim-van-hecke-alwhe/
Might take “a while” until you get reimbursed, an example from precision neurology

Precision Therapy
for Asthma and
Comorbidities
Is Asthma+X just another
phenotype in precision
therapy sense?

Asthma Comorbidities
Kwon et al. (2021): “Risk, Mechanisms and Implications of Asthma-
Associated Infectious and Inflammatory Multimorbidities (AIMs)
among Individuals With Asthma: a Systematic Review and a Case
Study” Cited by 5
Our prior work and the work of others have demonstrated that asthma
increases the risk of a broad range of both respiratory (e.g., pneumonia and
pertussis) and non-respiratory (e.g., zoster and appendicitis) infectious
diseases as well as inflammatory diseases (e.g., celiac disease and
myocardial infarction [MI]), suggesting the systemic disease nature of
asthma and its impact beyond the airways. We call these conditions
asthma-associated infectious and inflammatory multimorbidities
(AIMs). At present, little is known about why some people with asthma are
at high-risk of AIMs, and others are not, to the extent to which controlling
asthma reduces the risk of AIMs and which specific therapies mitigate
the risk of AIMs. These questions represent a significant knowledge gap in
asthma research and unmet needs in asthma care, because there are no
guidelines addressing the identification and management of AIMs.
This is a systematic review on the association of asthma with the risk of AIMs
and a case study to highlight that 1) AIMs are relatively under-recognized
conditions, but pose major health threats to people with asthma; 2) AIMs
provide insights into immunological and clinical features of asthma as a
systemic inflammatory disease beyond a solely chronic airway disease; and
3) it is time to recognize AIMs as a distinctive asthma phenotype in order
to advance asthma research and improve asthma care. An improved
understanding of AIMs and their underlying mechanisms will bring valuable
and new perspectives improving the practice, research, and public health
related to asthma.

Precision Therapy
and Drug Discovery
beyond Asthma
Hardly unique this disease
heterogeneity and need for
personalized medicine.
Big interest for Cancer and
Alzheimer’s for example.
Examples to provide some
broader context on the precision
asthma future.

Alzheimer’s Subtypes
Tijms et al. (2024): “Cerebrospinal fluid
proteomics in patients with Alzheimer’s disease
reveals five molecular subtypes with distinct
genetic risk profiles” Cited by 5
Alzheimer’s disease (AD) is heterogenous at the molecular level.
Understanding this heterogeneity is critical for AD drug
development. Here we define AD molecular subtypes using
mass spectrometry proteomics in cerebrospinal fluid (CSF),
based on 1,058 proteins, with different levels in individuals with
AD (n = 419) compared to controls (n = 187).
These AD subtypes had alterations in protein levels that were
associated with distinct molecular processes: subtype 1 was
characterized by proteins related to neuronal hyperplasticity;
subtype 2 by innate immune activation; subtype 3 by RNA
dysregulation; subtype 4 by choroid plexus dysfunction; and
subtype 5 by blood–brain barrier impairment. Each subtype was
related to specific AD genetic risk variants, for example, subtype
1 was enriched with TREM2 R47H.
Subtypes also differed in clinical outcomes, survival times
and anatomical patterns of brain atrophy. These results
indicate molecular heterogeneity in AD and highlight the need
for personalized medicine.

Precision Drug Development for Alzheimer’sa
Cheng et al. (2024): “Artificial intelligence and
open science in discovery of disease-modifying
medicines for Alzheimer’s disease”
The high failure rate of clinical trials in Alzheimer’s
disease (AD) and AD-related dementia (ADRD) is due to
a lack of understanding of the pathophysiology of
disease.
Recent accelerations in computing power and availability
of big data, including electronic health records (EHR) and
multiomics profiles, have converged to provide
opportunities for scientific discovery and treatment
development. Here, we review the potential utility of
applying AI approaches to big data for discovery of
disease-modifying medicines for AD/ADRD.
We illustrate how AI tools can be applied to the AD/ADRD
drug development pipeline through collaborative efforts
among neurologists, gerontologists, geneticists,
pharmacologists, medicinal chemists, and computational
scientists. AI and open data science expedite drug
discovery and development of disease-modifying
therapeutics for AD/ADRD and other neurodegenerative
diseases.

Precision Drug Development for Alzheimer’sb
Cheng et al. (2024): “Artificial intelligence and open science in
discovery of disease-modifying medicines for Alzheimer’s disease”

Precision Drug Development for Alzheimer’sc
Cheng et al. (2024): “Artificial intelligence and open
science in discovery of disease-modifying medicines for
Alzheimer’s disease”
A recent study presented network-based disease-progression-
specific drug repurposing for AD based on neuroimaging-derived
disease stages (Savva et al. 2022)
. Using three types of PET brain imaging
for microglial activation ([11C]PBR28)
, amyloid-b (Ab) ([18F]AZD4694)
, and tau
([18F]MK-6240)
, a recent study showed that the co-occurrence of Ab,
tau, and microglia abnormalities was the strongest predictor of
cognitive impairment (Pascoal et al. 2021)
.
This study highlighted that synergistically targeting microglial
abnormality, tau, and Ab endophenotypes could offer more
clinical benefits compared to targeting each endophenotypes
alone, supporting previous endophenotype-based drug
repurposing in AD (Fang et al. 2021)
.
Combination therapies via targeting molecular networks or
pathways derived from multi-omics data from individuals with well-
characterized proteinopathies and microglia abnormalities may
offer more effective treatment approaches compared to
monotherapies targeting one protein/microglia subtype. More
details about AI-based or network-based drug combination design
can be found in recent studies (94–96).
As drug development is a complex process involving many
steps, multi-modal machine learning tools can
significantly reduce the time and cost of drug
development. For instance, multi-modal machine learning
approaches (Venugopalan et al. 2021)
improve accuracy of patient
subphenotyping during clinical trial design by
assembling neuroimaging, genetic, and multi-omics
profiling data. With the help of deep learning, effective
representations can be learned for different data
modalities (Zhavoronkov et al. 2019)
, which can then be fused by
simple concatenation or more complicated nonlinear
transformation113 to perform downstream tasks such as
molecular design, PK and BBB property evaluation and
optimization, and robotics-based chemical synthesis,
which can greatly accelerate the drug discovery and
development process. If broadly applied, AI-based tools
will accelerate the development of disease-modifying
treatments for AD.

Precision Medicine for Parkinson’s
Höglinger et al. (2024): “A biological classification of
Parkinson's disease: the SynNeurGe research diagnostic
criteria”
With the hope that disease-modifying treatments could target the
molecular basis of Parkinson's disease, even before the onset of
symptoms, we propose a biologically based classification. Our
classification acknowledges the complexity and heterogeneity of
the disease by use of a three-component system (SynNeurGe):
presence or absence of pathological -synuclein (S) in tissues or
α
CSF; evidence of underlying neurodegeneration (N) defined by
neuroimaging procedures; and documentation of pathogenic gene
variants (G) that cause or strongly predispose to Parkinson's
disease. These three components are linked to a clinical
component (C), defined either by a single high-specificity clinical
feature or by multiple lower-specificity clinical features. The use of a
biological classification will enable advances in both basic and
clinical research, and move the field closer to the precision
medicine required to develop disease-modifying therapies. We
emphasise the initial application of these criteria exclusively for
research. We acknowledge its ethical implications, its limitations,
and the need for prospective validation in future studies.
Simuni et al. (2024): “A biological definition of neuronal
α-synuclein disease: towards an integrated staging
system for research”
Parkinson's disease and dementia with Lewy bodies are currently
defined by their clinical features, with -synuclein pathology as
α
the gold standard to establish the definitive diagnosis. We propose
that, given biomarker advances enabling accurate detection of
pathological -synuclein (ie, misfolded and aggregated) in CSF
α
using the seed amplification assay, it is time to redefine Parkinson's
disease and dementia with Lewy bodies as neuronal -synuclein
α
disease rather than as clinical syndromes.
Similar multimodal modeling needed as for Alzheimer’s

Cancer Precision Medicine #1
Petralia et al. (2024): “Pan-cancer proteogenomics
characterization of tumor immunity”
Despite the successes of immunotherapy in cancer treatment over
recent decades, less than <10%–20% cancer cases have
demonstrated durable responses from immune checkpoint
blockade. To enhance the efficacy of immunotherapies,
combination therapies suppressing multiple immune evasion
mechanisms are increasingly contemplated. To better understand
immune cell surveillance and diverse immune evasion responses in
tumor tissues, we comprehensively characterized the immune
landscape of more than 1,000 tumors across ten different cancers
using CPTAC pan-cancer proteogenomic data.
•Proteogenomics reveals seven immune subtypes
spanning 10 cancer types
•DNA alterations associate with immune subtypes and
affect proteomic profiles
•Kinase activation in immune subtypes suggests potential
therapeutic targets
•Digital pathology reveals infiltrating cells associated with
immune subtypes

Woodcock et al. (2024): “Genomic evolution shapes prostate
cancer disease type”
Tumor evolution is a dynamic process1 involving the accumulation of genetic
alterations that disrupt normal cellular processes, leading to pathological
phenotypes. While some cancers can be categorized into subtypes, often utilizing
pronounced genomic or transcriptomic differences, the evolutionary processes
that give rise to this variation are complex and not well understood. However, it
has been shown that the order of events in some hematological malignancies
can be related to prognosis and treatment susceptibility.
However, detailed investigations by ourselves11 and others have shown
substantial heterogeneity between tumors that presents challenges for simple
or consistent subtype assignments. Studies investigating evolutionary
differences between prostate cancer disease types by categorizing molecular
events as “early” or “late” have been shown to be informative in early-onset
and aggressive disease, and the temporal order of genetic alterations has also
been shown to be related to the ETS subtype.
• Tumors can be classified into evotypes according to evolutionary trajectory
• The evotype model unifies many previous molecular observations
Utility of evotype model in survival analysis
Adding order of molecular events (time) → evotypes

Gunjur et al. (March 2024): “A gut microbial signature for combination immune
checkpoint blockade across cancer types”
Immune checkpoint blockade (ICB) targeting programmed cell death protein 1 (PD-1)
and cytotoxic T lymphocyte protein 4 (CTLA-4) can induce remarkable, yet
unpredictable, responses across a variety of cancers. Studies suggest that there is a
relationship between a cancer patient’s gut microbiota composition and clinical
response to ICB; however, defining microbiome-based biomarkers that generalize
across cohorts has been challenging. This may relate to previous efforts quantifying
microbiota to species (or higher taxonomic rank) abundances, whereas microbial
functions are often strain specific.
Here, we performed deep shotgun metagenomic sequencing of baseline fecal samples
from a unique, richly annotated phase 2 trial cohort of patients with diverse rare cancers
treated with combination ICB (n = 106 discovery cohort). We demonstrate that strain-
resolved microbial abundances improve machine learning predictions of ICB
response and 12-month progression-free survival relative to models built using species-
rank quantifications or comprehensive pretreatment clinical factors. Through a meta-
analysis of gut metagenomes from a further six comparable studies (n = 364 validation
cohort), we found cross-cancer (and cross-country) validity of strain–response
signatures, but only when the training and test cohorts used concordant ICB regimens
(anti-PD-1 monotherapy or combination anti-PD-1 plus anti-CTLA-4). This suggests that
future development of gut microbiome diagnostics or therapeutics should be
tailored according to ICB treatment regimen rather than according to cancer type.
Measure ‘everything’ for ‘precision’

Foundation
model and
embedding
analysis extra

Foundation Models obviously here as well
Cui et al. (2024): “scGPT: toward building a foundation
model for single-cell multi-omics using generative AI”
https://github.com/bowang-lab/scGPT
Generative pretrained models have achieved remarkable
success in various domains such as language and computer
vision. Specifically, the combination of large-scale diverse
datasets and pretrained transformers has emerged as a
promising approach for developing foundation models. Drawing
parallels between language and cellular biology (in which texts
comprise words; similarly, cells are defined by genes), our study
probes the applicability of foundation models to advance cellular
biology and genetic research. Using burgeoning single-cell
sequencing data, we have constructed a foundation model for
single-cell biology, scGPT, based on a generative pretrained
transformer across a repository of over 33 million cells. Our
findings illustrate that scGPT effectively distills critical biological
insights concerning genes and cells. Through further adaptation of
transfer learning, scGPT can be optimized to achieve superior
performance across diverse downstream applications. This
includes tasks such as cell type annotation, multi-batch
integration, multi-omic integration, perturbation response
prediction and gene network inference.

Zero-shot for Protein Function prediction
Kulmanov et al. (2024): “Protein function prediction as
approximate semantic entailment”
https://github.com/bio-ontology-research-group/deepgo2
DeepGO-SE is a protein function prediction method that improves
the prediction performance for proteins by incorporating both
protein sequence features generated by a pretrained protein
language model, background knowledge from the GO and
interactions between proteins. Our results allow us to draw three
main conclusions: knowledge-enhanced machine learning
methods are now able to improve over methods that do not rely
on background knowledge; GO function prediction is best
formulated using a separate, hierarchical prediction approach;
and function prediction models based on ESM2 can now
generalize to largely unseen proteins.
In addition, DeepGO-SE is able to perform zero-shot predictions,
similar to DeepGOZero, and is faster to obtain predictions than
other methods that rely on multiple sequence alignments. This is
due to the fact that DeepGO-SE relies only on ESM2 embeddings,
which are faster to compute (Lin et al. 2023). Overall, the DeepGO-
SE model represents a significant improvement over existing
protein function prediction methods, providing a more accurate,
comprehensive and efficient approach.
@NatMachIntell @coolmaksat @CBRC_KAUST

UMAP Embeddings Problems
https://doi.org/10.1038/d41586-024-00568-w

Embeddings Extra #1
Javad Rahimikollu et al. (2024): “SLIDE: Significant
Latent Factor Interaction Discovery and Exploration
across biological domains” https://github.com/jishnu-lab/SLIDE
Modern multiomic technologies can generate deep multiscale
profiles. However, differences in data modalities, multicollinearity
of the data, and large numbers of irrelevant features make
analyses and integration of high-dimensional omic datasets
challenging. Here we present Significant Latent Factor Interaction
Discovery and Exploration (SLIDE), a first-in-class interpretable
machine learning technique for identifying significant interacting
latent factors underlying outcomes of interest from high-
dimensional omic datasets. SLIDE makes no assumptions
regarding data-generating mechanisms, comes with theoretical
guarantees regarding identifiability of the latent
factors/corresponding inference, and has rigorous false discovery
rate control. Using SLIDE on single-cell and spatial omic datasets,
we uncovered significant interacting latent factors underlying
a range of molecular, cellular and organismal phenotypes.
SLIDE outperforms/performs at least as well as a wide range of
state-of-the-art approaches, including other latent factor
approaches. More importantly, it provides biological inference
beyond prediction that other methods do not afford. Thus, SLIDE is
a versatile engine for biological discovery from modern
multiomic datasets.
UMAP visualization of the three stages
of clonal expansion (scRNA-seq and
TCR-seq data from NOD mice used to
infer mechanisms underlying clonal
expansion of CD4 T cells)

Embeddings Extra #2
Xia et al. (2024): “Statistical method scDEED
for detecting dubious 2D single-cell
embeddings and optimizing t-SNE and UMAP
hyperparameters” https://github.com/JSB-UCLA/scDEED
Popular methods such as t-distributed stochastic
neighbor embedding (t-SNE) and uniform manifold
approximation and projection (UMAP) are
commonly used for visualizing cell clusters;
however, it is well known that t-SNE and UMAP’s 2D
embeddings might not reliably inform the
similarities among cell clusters. Motivated by this
challenge, we present a statistical method, scDEED,
for detecting dubious cell embeddings output by
a 2D-embedding method. By calculating a
reliability score for every cell embedding based on
the similarity between the cell’s 2D-embedding
neighbors and pre-embedding neighbors, scDEED
identifies the cell embeddings with low reliability
scores as dubious and those with high reliability
scores as trustworthy. Moreover, by minimizing the
number of dubious cell embeddings, scDEED
provides intuitive guidance for optimizing the
hyperparameters of an embedding method.

Federated Learning for pooling data
Zhou et al. (2024): “PPML-Omics: A privacy-preserving federated machine
learning method protects patients’ privacy in omic data”
Modern machine learning models toward various tasks with omic data analysis
give rise to threats of privacy leakage of patients involved in those datasets.
Here, we proposed a secure and privacy-preserving machine learning method
(PPML-Omics) by designing a decentralized differential private federated
learning algorithm. We applied PPML-Omics to analyze data from three
sequencing technologies and addressed the privacy concern in three major
tasks of omic data under three representative deep learning models. We
examined privacy breaches in depth through privacy attack experiments and
demonstrated that PPML-Omics could protect patients’ privacy.
In each of these applications, PPML-Omics was able to outperform methods of
comparison under the same level of privacy guarantee, demonstrating the
versatility of the method in simultaneously balancing the privacy-preserving
capability and utility in omic data analysis. Furthermore, we gave the
theoretical proof of the privacy-preserving capability of PPML-Omics,
suggesting the first mathematically guaranteed method with robust and
generalizable empirical performance in protecting patients’ privacy in omic
data.

When does
‘Precision’
become clinical
reality then?

What will be the cost and ease of running omics analysis?
To get a sense of how powerful digital transformation is on genomics, in 2005, a single sequencing run could only
produce one gigabase (one billion base pairs) of data and cost $14 million. By late 2014, that rate had climbed to 1.8
terabases of data (1,800x increase) while the cost has come down to $4,000 (3,500x decrease). And just this year,
Illumina’s HiSeq X machine broke $1,000. d3.harvard.edu
“Omics” getting cheaper, but drug development more expensive
Tim Fitzpatrick Co-Founder & CEO at IKONA More recently,
#biotech companies are now using AI to try to break Eroom's law—
and they may be closer than you think. Here are a few companies
using #AI to tackle speed, cost, and accuracy across dozens of
therapeutic areas:
Exscientia (founded 2012, 500 ppl); Insilico Medicine (2014, 250 ppl);
Schrödinger (1990, 900 ppl); Recursion (2013, 570 ppl); Genentech
(1976, 20K ppl); Atomwise (2012, 75 ppl); Relay Therapeutics (2016,;
340 ppl); Molecule.one (2016, 26 ppl); CellVoyant (2021, 12 ppl);
CHARM Therapeutics (2021, 59 ppl); Isomorphic Labs (2021, 82 ppl);
Cradle (2021, 37 ppl); AQEMIA (2019, 50 ppl); Iktos (2016, 66 ppl);
Multiomic Health (2021, 18 ppl)

ML has its generalization issues #1
https://doi.org/10.1038/d41586-024-00094-9
https://doi.org/10.1126/science.adm9218
Machine learning models have been heralded as the
tools to accelerate precision medicine by sifting through
large amounts of complex data to pinpoint the genetic,
sociodemographic, or biological markers that predict
the right treatment for the right person at the right time.
However, the initial enthusiasm for these advanced
predictive tools is now facing a sobering reality check.
Chekroud et al. (2024, Science): “Illusory generalizability of clinical
prediction models”
It is widely hoped that statistical models can improve decision-making related to medical
treatments. Because of the cost and scarcity of medical outcomes data, this hope is typically
based on investigators observing a model’s success in one or two datasets or clinical
contexts. We scrutinized this optimism by examining how well a machine learning model
performed across several independent clinical trials of antipsychotic medication for schizophrenia.
Models predicted patient outcomes with high accuracy within the trial in which the model was
developed but performed no better than chance when applied out-of-sample. Pooling data
across trials to predict outcomes in the trial left out did not improve predictions. These results
suggest that models predicting treatment outcomes in schizophrenia are highly context-
dependent and may have limited generalizability.

ML has its generalization issues #2
Lasko et al. (2024): “Why do probabilistic clinical models fail to transport
between sites”
The rising popularity of artificial intelligence in healthcare is highlighting the problem that
a computational model achieving super-human clinical performance at its training
sites may perform substantially worse at new sites. In this perspective, we argue that
we should typically expect this failure to transport, and we present common sources
for it, divided into those under the control of the experimenter and those inherent to the
clinical data-generating process. Of the inherent sources we look a little deeper into
site-specific clinical practices that can affect the data distribution, and propose a
potential solution intended to isolate the imprint of those practices on the data from the
patterns of disease cause and effect that are the usual target of probabilistic clinical
models.
Failure to transport was recognized as a problem by the earliest medical AI pioneers. In 1961, Homer
Warner implemented the first probabilistic model of symptoms and disease (Warner et al. 1961) using
a Naïve Bayes method (Ledley and Lusted 1959) with local conditional probabilities, which then failed
on external data. A follow-up by Bruce and Yarnall (1966) using data from three sites noted similar
failure to transport between sites, which they attributed to differences in conditional probabilities. A
decade later, Alvan Feinstein argued that the very idea of probabilistic diagnosis was fatally
flawed, on the grounds that the observational accuracy, the prevalence, and even the definitions of
collected clinical observations varied across sites (Feinstein 1973). A decade after that, in the
inaugural issue of Medical Decision Making, Tim de Dombal doubted ever being able to design a
probabilistic diagnosis engine with data from one site that worked at others, because large-
scale surveys demonstrated that disease prevalence and presentation vary dramatically across
locations (de Dombal et al. 1981).
Separating the site-specific Process Model P(C,E|
O,T,S) from the stable Disease Model P(Z,Y|C,E).

Polygenic Risk Score
https://erictopol.substack.com/p/polygenic-risk-scores-ready-for-prime
eMerge, asthma
Lennon et al. (2024): “Selection, optimization and validation of ten chronic disease
polygenic risk scores for clinical implementation in diverse US populations”

Future Genomics Prospects
Tuuli Lappalainen et al. (2024): “Genetic and
molecular architecture of complex traits”
Studies of molecular and cellular effects of genetic variants have
provided insights into biological processes underlying disease. Many
outstanding questions remain, but the field is well poised for
groundbreaking discoveries as it increases the use of genetic data
to understand both the history of our species and its applications to
improve human health.
Genetic association studies (GWAS) are finally starting to cover the full
spectrum of different types of variants across the frequency spectrum. The
integration of population genetics and statistical genetics provides a rich
opportunity for improved mapping of genetic architecture of complex traits. As
the number of robustly identified genetic associations has exploded, the
challenge of their functional interpretation has become a central question in
the field, now tackled with a combination of tools expanding beyond genetics to
molecular computational biology.
Beyond generating more data, advances are likely to come from the
synthesis of insights across the disciplines of quantitative, molecular,
population, and epidemiological genetics. Understanding the causes and
consequences of disease architecture (widespread polygenicity and pleiotropy,
in particular) will require advances in quantitative genetics to incorporate
parameters of natural selection coupled with advances in genetic
epidemiology to understand the relevant environmental contexts and risk
factors shared across traits.

Link genomic data with clinical data
Sosinsky et al. (2024): “Insights for precision oncology from
the integration of genomic and clinical data of 13,880 tumors
from the 100,000 Genomes Cancer Programme”
Genomics England, alongside NHS England, analyzed WGS data from 13,880 solid
tumors spanning 33 cancer types, integrating genomic data with real-world
treatment and outcome data, within a secure Research Environment. Incidence
of somatic mutations in genes recommended for standard-of-care testing varied
across cancer types. For instance, in glioblastoma multiforme, small variants were
present in 94% of cases and copy number aberrations in at least one gene in 58%
of cases, while sarcoma demonstrated the highest occurrence of actionable
structural variants (13%). Homologous recombination deficiency was identified in
40% of high-grade serous ovarian cancer cases with 30% linked to pathogenic
germline variants, highlighting the value of combined somatic and germline
analysis.
The linkage of WGS and longitudinal life course clinical data allowed the
assessment of treatment outcomes for patients stratified according to
pangenomic markers. Our findings demonstrate the utility of linking genomic
and real-world clinical data to enable survival analysis to identify cancer genes
that affect prognosis and advance our understanding of how cancer genomics
impacts patient outcomes.
For Precision Oncology

Epigenetics
Forno and Celedón (2019): “Epigenomics and Transcriptomics in the
Prediction and Diagnosis of Childhood Asthma: Are We There Yet?”
Cited by 29
Most traditional approaches have been based on garnering clinical
evidence, such as risk factors and exposures. Given the high heritability
of asthma, more recent approaches have looked at genetic
polymorphisms as potential “risk factors.” However, genetic variants
explain only a small proportion of asthma risk, and have been less
than optimal at predicting risk for individual subjects. Epigenomic
studies offer significant advantages over previous approaches.
Epigenetic regulation is highly tissue-specific, and can induce both
short- and long-term changes in gene expression. Such changes can
start in utero, can vary throughout the life span, and in some instances
can be passed on from one generation to another. Most importantly, the
epigenome can be modified by environmental factors and exposures,
and thus epigenetic and transcriptomic profiling may yield the most
accurate risk estimates for a given patient by incorporating
environmental (and treatment) effects throughout the lifespan. Here we
will review the most recent advances in the use of epigenetic and
transcriptomic analysis for the early diagnosis of asthma and atopy, as
well as challenges and future directions in the field as it moves forward.
We will particularly focus on DNA methylation, the most studied
mechanism of epigenetic regulation.
Papamichael and Katsardis (2024): “Nutrient intake,
epigenetics, and asthma”
Asthma, a chronic inflammatory respiratory disorder, is
associated with a considerable burden and
necessitates an improved understanding of the
underlying disease mechanisms. Epigenetics, a
budding science defined as the study of mitotically
and/or meiotically heritable but reversible changes of
a phenotype that do not cause alterations in DNA
sequence, may explain how genetic susceptibility
and exposure to environmental stimuli, including
diet, could interact in defining respiratory diseases. It
is possible that adaptation to the environment could
activate alterations in gene expression, predisposing
the individual to asthma. The purpose of this narrative
review is to provide an overview of the latest scientific
evidence on (1) the role of epigenetics in asthma
inception, (2) dietary influences on gene
expression/DNA methylation in relation to asthma
etiology, and (3) uncover potential epigenetic traits
and possible molecular targets for the development
of future primary preventive strategies and
personalized treatment of this disease.

Single-cell biology
Cormac Sheridan (Feb 2024): “Can single-cell biology realize
the promise of precision medicine?”
Biology’s quiet revolution is underway, as single-cell tools fuel
the next-wave of drug discoveries and promise to match
therapies to the individual.
Cellarity’s single-cell-based approach to drug discovery is at
the heart of a recent agreement with Novo Nordisk to develop
new drugs for a form of chronic fatty liver disease associated
with metabolic dysfunction. The hope is that this shift away
from molecular targets and toward identifying the
underlying cellular dysfunction will offer deep insights into
the biology of the condition, which will, in turn, lead to improved
therapies. Just as the microscope enabled the development of
microbiology in the seventeenth century, high-resolution,
single-cell methods are opening up whole new biological vistas
At this point, single-cell analysis is nowhere near routine
clinical practice, but a handful of pioneering studies
points to its potential in realizing the largely unattained
goal of precision medicine — that is, accurately
matching an individual patient to the therapies from
which they are most likely to benefit. “Clinically, I would
say, we’re at the very early stages,” says Joseph Powell, of
the Garvan Institute of Medical Research in Darlinghurst,
Australia.
Single-cell analysis is not confined to’omic analytes,
either. Exscientia, of Oxford, UK, is throwing its weight
behind another foundational technology: high-
throughput single-cell imaging. When combined with
machine learning, it can yield important insights into the
status of a cell exposed ex vivo to a drug. This approach
to drug discovery is inherently unbiased — it does not
require prior knowledge of the genomic profile of a
cancer to recommend a particular course of action but
relies instead on artificial intelligence-driven image
analysis of cells following their exposure. The company
gained the technology through its acquisition of Vienna-
based Allcyte in 2021.

ML UX for
Clinical Practice
As in how to present the
information, and try to have
the clinicians engaged for
human-AI collaboration (see
e.g. Doshi-Velez at NeurIPS
2023)

Digital Tools in Asthma
Mosnaim et al. (2021): “Digital Health Technology in Asthma: A
Comprehensive Scoping Review”
Of 871 articles identified, 121 were evaluated to explore
intervention characteristics, the perception and
acceptability of digital interventions to patients and
physicians, and effects on asthma outcomes. Interventions
were categorized by their level of interactivity with the patient.
Digital interventions were generally positively perceived by
patients and physicians. Implementation was considered
feasible, with certain preferences for design and features
important to drive use.
Digital health interventions show substantial promise for
asthma disease monitoring and personalization of
treatment. To be successful, future interventions will need to
include both inhaler device and software elements,
combining accurate measurement of clinical parameters
with careful consideration of ease of use, personalization, and
patient engagement aspects.

Asthma Management Tool
Zheng et al. (2024): “Clinical Needs Assessment of a Machine
Learning–Based Asthma Management Tool: User-Centered
Design Approach”
Personalized asthma management depends on a clinician’s ability
to efficiently review patient’s data and make timely clinical
decisions. While machine learning (ML) and clinical decision support
tools are well-positioned as potential solutions, the translation of such
frameworks requires that barriers to implementation be addressed in
the formative research stages.
The : Asthma Guidance and Prediction System (A-GPS) tool is an ML-
based CDS tool accessible from “within” the EHR workflow. It aims to
summarize all asthma-related context information extracted from the
EHR on 1 screen page [Seol et al. 2020, Seol et al. 2021]. The tool will be
embedded with a functional component of the asthma exacerbation
(AE) risk model, which applies ML algorithms to predict a patient’s
risk of exacerbation in 1 year [Overgaard et al. 2022]. Moreover, the
reason for the development of the AE risk model in this context was to
provide supplemental information to improve the care of patients with
asthma, not replace the expertise of clinicians.

Co-ideating with stakeholders
Leow et al. (2023): “Innovation workshop using design thinking
framework and involving stakeholders to co-create ideas for
management of asthma”
A one-day workshop was organised and conducted in Singapore by
SingHealth Polyclinics to discuss the challenges and enablers of good
asthma care and ideate innovations to address the issues discussed. The
workshop was conceptualised based on the Stanford’s d: school
Design Thinking Process: 1. empathise, 2. define, 3. ideate, 4. prototype,
and 5. test, focussing on the first three stages. Empathise stage was
executed by having two patients share their challenges and enablers
of good asthma care using photovoice. Define and ideate stage were
accomplished through the multidisciplinary team discussion, with the
patient going to every group to allow them to seek clarifications and
opinions on ideas.
More than 90% of participants strongly agreed or agreed that they
could generate ideas to improving asthma care, the workshop helped
drive innovation, and the use of photovoice helped them empathise with
patients challenges. A design thinking framework can be used for
innovation workshops. Photovoice is a useful method for understanding
the problems faced by patients. A multidisciplinary team format with
patient involvement was highly favoured.

Check other fields with more digital tools
https://phti.com/assessment/digital-diabetes-management-tools/

Respiratory
Disease
Conferences

ATS International Conference
https://conference.thoracic.org/

ERS International Congress
https://www.ersnet.org/congress-and-events/congress/
Vol 62, Issue suppl 67, Page range
ERS International Congress 2023 abstracts

Precision Medicine for personalized treatment of asthma

More Related Content

What's hot

Similar to Precision Medicine for personalized treatment of asthma

More from PetteriTeikariPhD

Recently uploaded

Precision Medicine for personalized treatment of asthma