All-domain Anomaly Resolution Office U.S. Department of Defense (U) Case: “Eg...
Nick Tatonetti's presentation on Systems Pharmacology at AMIA 2015
1. Systems pharmacology for
drug safety
November 15th, 2015
Nicholas P. Tatonetti, PhD
Herbert Irving Assistant Professor of Biomedical Informatics
Columbia University
2. Observation is the starting point
of biological discovery
• Charles Darwin observed relationship
between geography and phenotype
• William McBride & Widukind Lenz
observed association between
thalidamide use and birth defects
3. The tools of observation are
advancing
• Human senses
• sight, touch, hearing, smell, taste
• Mechanical augmentation
• binoculars, telescopes, microscopes,
microphones
• Chemical and Biological augmentations
• chemical screening, microarrays, high
throughput sequencing technology
• What’s next?
Bytes to KB
Megabytes to
Terabytes
4. The tools of observation are
advancing
• Human senses
• sight, touch, hearing, smell, taste
• Mechanical augmentation
• binoculars, telescopes, microscopes,
microphones
• Chemical and Biological augmentations
• chemical screening, microarrays, high
throughput sequencing technology
• What’s next?
Bytes to KB
Megabytes to
Terabytes
5. Technological Augmentation
• Tech companies are becoming really good at
observing (and recording) the moments of life
• Facebook
• Google
• Apple (iCloud)
• 2015, the year of the zetabyte
• 1 zetabyte = 1,000 exabytes = 1 billion terabytes
6. Your doctor is observing you
like never before
>99% of Hospitals have Electronic Health Records
7. Your doctor is observing you
like never before
>60% of ALL Physicians
9. Observation analysis in a petabyte world
• Darwin, McBride, and Lenz were working with
kilobytes of data
• Today’s scientists are observing terabytes and
petabytes of data
• The human mind simply cannot make sense of that
much information
• Data mining is about making the tools of data
analysis (“hypothesis generation”) catch up to the
tools of observation
12. Databases of drug effects are
confounded
• Most drug side effects are only discovered after drugs hit
the market using observational data
• This leads to high false positive and false negative rates
when using EHR and adverse event data to find side effects
A
B
A
13. MWAS
Ryan et al. CPT: Pharmacometrics & Systems Pharmacology (2013)
False positives
False negatives
FN: Estradiol, Desipramine
Sensitivity: 67% 70%
Specificity: 60% 60%
Myocardial Infarction
Medication-wide association study
15. Systems pharmacology
• Integration of physiological, biochemical, genomic
data to analyze drug actions and side effects in the
context of the interactome
• Key method: network analysis
• Nodes = proteins and small molecules
• Edges = interactions
(aka systems pharmacology)
16. MADSS
• Use network analysis to build
AE neighborhoods: a subset
of the interactome surrounding
AE “seed” proteins
• Score each protein on
connectivity to seeds
• Overarching hypothesis:
drugs targeting proteins within
an AE neighborhood more
likely to be involved in
mediating that AE
Modular Assembly of Drug Safety Subnetworks
Lorberbaum, et al. Clin. Pharmacol. Ther. (2015)
Protein
Interaction
Seed protein
Adverse event
Drug known to cause AE
Drug predicted to cause AE
17. • For each AE, use four adapted pairwise
connectivity metrics to score every protein in
interactome on its connectivity to the seed set
• Mean first passage time (MFPT)
• Betweenness centrality (BC)
• Shared neighbors (SN)
• Inverse shortest path (ISP)
Building AE neighborhoods
Lorberbaum, et al. Clin. Pharmacol. Ther. (2015)
20. Serotonin agonists (triptans): serotonin
receptor activation can lead to vaso-
constriction and increased synthesis of
IL-6 (MI seed) in vascular smooth muscle
21. Can we use these molecular data to
predict results of clinical trials or
post-market surveillance?
22. Evaluating MADSS:
Drug safety gold standard
• Gold standard for 4 AEs created using systematic literature
review and natural language processing of structured
product labeling
GI Bleeding (73)
24 positives
49 negatives
Myocardial Infarct (73)
33 positives
40 negatives
Liver Failure (89)
63 positives
26 negatives
Kidney Failure (49)
19 positives
30 negatives
23. MWAS
Ryan et al. CPT: Pharmacometrics & Systems Pharmacology (2013)
False positives
False negatives
FN: Estradiol, Desipramine
Sensitivity: 67% 70%
Specificity: 60% 60%
Myocardial Infarction
Medication-wide association study
24. Evaluating MADSS:
Subnetwork (SubNet) models
• Trained SubNet model for each AE individually
using connectivity scores as features
• Evaluated MWAS alone, SubNet alone, MWAS
+SubNet
25. GI Bleeding Liver Failure
Kidney FailureMyocardial Infarct
MWAS + SubNet
Systems Pharmacology
Alone (SubNet)
Statistics Alone (MWAS)
Comparing network biology to post-market analysis
Lorberbaum, et al. Clin. Pharmacol. Ther. (2015)
26. MWAS+SubNet outperforms either model alone
Drug MWAS SubNet Both
Desipramine 70% 85% 100%
Darbepoetin
Alfa
49% 73% 100%
Estradiol 67% 52% 75%
Frovatriptan 42% 64% 75%
Imipramine 64% 58% 67%
Myocardial Infarct
Sensitivity Specificity
MWAS SubNet Both
60% 74% 100%
80% 86% 100%
60% 89% 100%
89% 86% 100%
71% 89% 100%
Lorberbaum, et al. Clin. Pharmacol. Ther. (2015)
27. Observational analysis is the fuel of
scientific discovery
• Data-mining has the potential to generate billions of
hypotheses we could not have conceived of
• However, like all good hypotheses, these must be
rigorously tested
• Systems pharmacology reveals the molecular
hypothesis of drug side effects enabling
experimental validation
28. 28
tatonettilab.org
nick.tatonetti@columbia.edu
@nicktatonetti
Current Lab Members
Robert Moskovitch, PhD
Rami Vanguri, PhD
Alexandra Jacunski
Tal Lorberbaum**
Mary Boland
Joseph Romano
Yun Hao
Phyllis Thangaraj
Alexandre Yahi
Collaborators
Brent Stockwell, PhD
George Hripcsak, MD, MS
Ziad Ali, MD, DPhil
Santiago Vilar, PhD
Konrad Karczewski, PhD (Broad/MGH)
Joel Dudley, PhD (Mount Sinai)
Patrick Ryan, PhD (OHDSI)
Eric Horvitz (Microsoft Research)
Ryen White (Microsoft Research)
Russ Altman (Stanford)
Funding
NIGMS R01GM107145
Herbert Irving Fellowship
NCI P30CA013696
NIMH R03MH103957
PhRMA Foundation
AstraZeneca
Thank you