Public Domain 1 BEL at the Heart of Pfizer’s Systems Biology Infrastructure OpenBEL Workshop April 29, 2014 Carol Scalice, Business Partner, Systems Biology.

Public Domain

1

BEL at the Heart of Pfizer’s Systems Biology Infrastructure

OpenBEL Workshop

April 29, 2014

Carol Scalice,

Business Partner, Systems Biology

Public Domain

The views presented in this talk do not necessarily represent the views or opinions of Pfizer, Inc. These materials do not represent a guarantee and Pfizer, Inc. is not liable for any attempts to reuse these materials or the concepts contained herein.

2

Public Domain

Outline

What have we done with BEL?

What are we working on now and where are we headed?

Q&A

Causal Reasoning with Transcriptional Data(Classic/Omics CRE)

• Input is set of of up- and down-regulated transcripts (e.g., between healthy and disease state).

• Output is set of hypotheses of potential molecular causes consistent with input.

• Basic concept outlined in Pollard et al. (2004).

= potential causal hypothesis

Ziemek, D. (2014) Interpreting genetics and transcriptomics data using the Causal Reasoning Engine.

Bayes CRE: Contextual Analysis of Hypotheses

• Consider PPARG’s function in immunology and adipogenesis.

• Resulting hypotheses will include qualification of applicable context.

Zarringhalam et al,,Bioinformatics 2013

Ziemek, D. (2014) Interpreting genetics and transcriptomics data using the Causal Reasoning Engine.

Public Domain

Causal Interaction Query

6

Public Domain

Causal Interaction Query

7

Public Domain

Proprietary Experimental Results

8

[compound] a(PFE-c:nnnnnnn) => g(EGID:11200)

Activation

InhibitionAgonist

BindingIC50 EC50

Antagonist

Public Domain

Outline

What have we done with BEL?

What are we working on now and where are we headed?

Q&A

Precision Medicine: can we find the subset of patients who will respond to a particular drug?

• Most disease are heterogeneous. Can we find better subclasses?

• Given genetics, transcriptional and clinical data can we predict who will respond to treatment?

• Data generation is becoming cheaper and cheaper…

Ziemek, D. (2014) Interpreting genetics and transcriptomics data using the Causal Reasoning Engine.

Use causal knowledgebase to formalize that idea..and use upstream regulators as variables.

• For each patient, compute hypothesis profile as output.

• Each hypothesis gets a posterior probability between 0 and 1 as score.

• Followed by L1-regularized regression.

Hypothesis probability

LPS + 0.99

ERBB2 + 0.6

…

Patient GSM364633(Non-Responder)

Gene Value

t(FOXO1) +1

t(IRF7) -1

…

Patient GSM364633(Non-Responder)

Ziemek, D. (2014) Interpreting genetics and transcriptomics data using the Causal Reasoning Engine.

Public Domain

12

Image courtesy of Dexter Pratt, Ideker Labs, UCSD

Public Domain

13

NDEx

NDEx

BEL Compiler

Public Domain

14

XGMML

XGMML

NDExXGMML

XGMML

Public Domain

Computational Biology for Drug Discovery

15

Public Domain

Questions

Public Domain

PublicationsCausal Reasoning on Biological Networks: Interpreting transcriptional changesL Chindelevitch, D Ziemek, A Enayetallah, R Randhawa, B Sidders, C Brockel, E Huang

Assessing Statistical Significance in Causal GraphsL. Chindelevitch, P. Loh, A. Enayetallah, B. Berger and D. Ziemek

Modeling the Mechanism of Action of a DGAT Inhibitor Using a Causal Reasoning PlatformA. Enayetallah, D. Ziemek, M. Leininger, R. Randhawa, et al.

Novel Pancreatic Endocrine Maturation Pathways Identified by Genomic Profiling and Causal ReasoningA Gutteridge, JM Rukstalis, D Ziemek, M Tié, L Ji, et al.

Genes contributing to pain sensitivity in the normal population: an exome sequencing studyFMK Williams, S Scollen, D Cao, Y Memari, CL Hyde, B Zhang, B Sidders, D Ziemek, et al.

Molecular causes of transcriptional response: a Bayesian prior knowledge approachK Zarringhalam, A Enayetallah, A Gutteridge, B Sidders, D Ziemek