Presentation at the 2018 American Conference on Pharmacometrics

1. July
Dynamic models for personalized QSP
How models can help us explore big data
Ioannis (Yannis) P. Androulakis
Biomedical Engineering and
Chemical & Biochemical Engineering, Rutgers University
Department of Surgery, Rutgers-RWJ Medical School
American Conference on Pharmacometrics
October 20, 2018

2. July
Some nomenclature
Big data: the study, and applications, of data sets that are too complex for traditional
data-processing application software to adequately deal with (Wikipedia)
Models: [something] that requires computational resources [used to] to study the
behavior of a complex system (paraphrasing Wikipedia)
Personalized medicine: steering the right patients to the right drug at the right dose
at the right time (Hamburg and Collins, NEJM, 363(4):301, 2010)
PM is much more complicated that this – it should account for family history,
behaviors, environment, healthcare system etc., but for this discussion it suffices

3. July
The big picture
Trying to figure out
1. If there is a problem
2. What caused the problem
3. How to fix the problem
If … By and large we are driven by the concept of “the” biomarker, i.e., [an] objective
indication of [the] medical state observed from outside the patient (Curr Opin HIV
AIDS, 5(6):463, 2010)
What … We then establish a functional relation between the biomarker and a likely
deregulation of a mechanism
How … Finally, we attempt to manipulate the mechanism using a substance i.e., drug,
the can induce the desired change on the mechanism.

4. July
The big picture
Trying to figure out
1. If there is a problem
2. What caused the problem
3. How to fix the problem
If … By and large we are driven by the concept of “the” biomarker, i.e., [an] objective
indication of [the] medical state observed from outside the patient (Curr Opin HIV
AIDS, 5(6):463, 2010)
What … We then establish a functional relation between the biomarker and a likely
deregulation of a mechanism
How … Finally, we attempt to manipulate the mechanism using a substance i.e., drug,
the can induce the desired change on the mechanism.
http://www.animalresearch.info/en/dru
g-development/drug-
prescriptions/simvastatin/

5. July
On genes, drugs and models
Quantitative/Systems Pharmacology models are the glue that brings together the
biomarker(s), the mechanism(s) and the drug(s)
The interpretation of the model is what helps us rationalize data and infer system
properties
Figure adapted from Bill Jusko

6. July
Promise of big data
Big data is not (much) more of the same: types;
dimensions within a scale; scales; conditions;
subjects; everything
Big data makes the problem (much more)
multidimensional, in more ways than one
But … big data are “too complex” to analyze, so the
question is how to “upgrade the information
content of big data”
Hypothesis: computational models, and their
interpretation, act as the integrators of the
information captured by big data

7. July
Physiological mechanisms act as big data
integrators: the case of the amazing HRV
Physiological alarms do not respond to a single signal but rather have ways of
integrating distributed information
Heart Rate Variability: The amazing biomarker indicative of health
Heart Intl, 11(1):32, 2016

8. July
Subtle changes in the cardiac output precede
biochemical biomarkers: harnessing the information
Cardiac output (beat-to-beat variability) integrates diverse information describing the
state of the host and generates a signal which when interpreted computationally
produces a biomarker which precedes biochemical detection of sepsis

9. July
Pathways/Networks as functionally related entities –
The population static network
Perturbations of individual elements may
lack explicit structure, but when taken
together in the context of a functional
groupings become consistent
Langley et al., Sci Transl Med, 2013 Kamisoglu et al., Crit. Care, 3(19):71, 2015

10. July
Pathway dynamics as an emergent property –
The population dynamic network
Network dynamics is an emergent property, resulting from component interactions
Acevedo et al., (under review)
Acute MPL dosing
Chronic MPL dosing
Different components connected through the dynamics of the pathway

11. July
Personalized perturbation of pathways -
The personalized static network
Patients with the same disease/response manifest highly heterogeneous component
responses, nevertheless pointing towards common functional disruption
Different genes could correspond to tightly regulated checkpoints within the pathway.
Deviation from homeostasis results in consistent disease-specific pathway activation
despite “inconsistent” component changes
Menche et al., NPJ Syst Biol Appl, 3(10) 2017

12. July
Response dynamics of personalized networks –
The personalized dynamic network
Fey et al., Sci Signaling, 8(408):1, 2015

13. July
From genotypic plasticity to phenotypic similarity to
personalized response diversity
Entrainment and
adaptation varies
Response to acute stress Long term adaptation to stressRao and Androulakis, Endocrinology, 158(11):4017, 2017

14. July
Some final thoughts
Personalized medicine implicitly requires big data
• large number of individuals
• the differentiation at the individual level requires increased resolution
Modeling is a “state of mind” that can enable the successful application of big data
at the level of an individual patient

15. July
androulakis ip
www.ipandro.com