EP 160: Artificial Intelligence, GWAS in Drug Discovery, and Career Insights with Dr. Eric Fauman, Executive Director and Head of Computational Biology in the Internal Medicine Research Unit at Pfizer

The Genetics Podcast

0:00 Introduction

1:30 The power of social media: How Eric published 10 papers based on ideas that he discussed on Twitter

5:50 Explanation of The Table of Everything, an internal database at Pfizer that catalogs nearly 20,000 human genes and their associated diseases and traits

13:20 How Eric’s team works to correlate genome-wide association study (GWAS) results to real biological phenotypes and outcomes

18:10 Introduction to protein quantitative trait locus (PQTL), including its importance in biological and genetic data

25:10 Examining the evolving bottlenecks in drug development and the challenges of validating genetic targets

28:30 Navigating the gap between genetic hits and biological understanding, and how AI or functional studies could bridge this in target discovery

32:20 Linus Pauling's mentorship of Eric and how he might react to AlphaFold2’s breakthroughs in structural biology

35:15 Eric's take on using AI and how he's experimenting with it on trusted datasets

41:00 An introduction to Mendelian randomization, as well as its strengths and limitations

47:00 How Eric uses the TOP Model (Talent, Opportunity, and Passion) to guide this career choices and path

52:00 Diversity and collaboration in genetics research and implementation

55:00 Closing remarks

Resources mentioned throughout the episode:

Mendelian Randomization with Proxy Biomarkers

Paper: Mendelian randomisation with proxy exposures: challenges and opportunities, I Rahu, R Tambets, EB Fauman, Kaur Alasoo (2024)

Explores proxy biomarkers as a method to assess in vivo activity of a protein target.

Trait Colocalization and Causal Genes

Paper: Integrative analysis of metabolite GWAS illuminates the molecular basis of pleiotropy and genetic correlation, CJ Smith, N Sinnott-Armstrong, A Cichońska, H Julkunen, EB Fauman, Jonathon Pritchard, Elife 11, e79348

Demonstrates how traits with opposing effects on a genetic variant may suggest a causal gene sits between them

Metabolite Profiling in Human Knockouts

Paper: McGregor TL, Hunt KA, Yee E, et al. Characterising a healthy adult with a rare HAO1 knockout to support a therapeutic strategy for primary hyperoxaluria, Elife. 2020;9. Published 2020 Mar 24.

Community Workshop on Effector Gene Standards

Presentation: Watch on YouTube

TOP Model for Career Guidance

Article: Grab the Helm: How to Take Charge of Your Purpose, Passion, Progress

The Table of Everything

Overview: Read more on Pfizer’s site

UK Biobank Protein QTL Study

Paper: Sun, B.B., Chiou, J., Traylor, M. et al. Plasma proteomic associations with genetics and health in the UK Biobank,Nature, 622, 329–338 (2023).

Eric’s First GWAS Contribution

Paper: Shin SY, Fauman EB, Petersen AK, et al. An atlas of genetic influences on human blood metabolites, Nat Genet.2014;46(6):543-550.

Every Gene Ever Annotated (EGEA)

Public Resource: View annotations on GitHub

Nine reasons not to use eQTLs to identify causal genes from GWAS:

Random Sequences Can Create Regulatory Elements

• “~83% of random promoter sequences yielded measurable expression” - de Boer CG, Nat Biotechnol, 2020
• “Recently evolved enhancers are formed predominantly by exaptation of ancestral DNA” - Villar D, Cell, 2015
• “Extensive co-regulation of neighboring genes complicates the use of eQTLs in target gene prioritization” - Tambets R, et al., HGG Adv., 2024

Enhancer Variants and Buffering in Important Genes

• “eQTLs at GWAS loci are more likely to point to genes with low enhancer redundancy not associated with disease” - Wang X, Goldstein DB, Am J Hum Genet., 2020
• “GWAS and eQTL studies are systematically biased toward different types of variants” - Mostafavi H, et al., Nat Genet., 2023
• “CNVs are buffered by post-transcriptional regulation in 23%-33% of proteins significantly enriched in protein complex members” - Gonçalves E, et al., Cell Systems, 2017

eQTL Data Limitations vs. Proximity Information

• “cis-eQTL target genes are relatively poor indicators of ‘true positive’ causal genes” - Stacey D, et al., NAR., 2018
• “When molecular QTL colocalization evidence was removed, we saw similar classification results” - Mountjoy E, et al., Nat Genet., 2021
• “Key predictive features included coding or transcript-altering SNVs, distance to gene, and open chromatin-based metrics” - Forgetta V, et al., Hum Genet., 2022