When I joined Pfizer’s Internal Medicine Research Unit, genetics-based target discovery relied on one-off analyses run by individual scientists on a high-performance computing (HPC) cluster. Nothing reusable connected genome-wide association study (GWAS) evidence to functional genomics to a target nomination. Over four years I led the team that built that pipeline. It combined human genetic evidence (GWAS, exome-wide association studies, colocalization, and Mendelian randomization) with functional genomics, including single-cell chromatin accessibility, expression and protein quantitative trait locus (eQTL and pQTL) data, and deep learning predictions of variant function, to find and rank new targets with genetic support for efficacy and selectivity.
I also led the move of Pfizer’s genomics analysis to Amazon Web Services (AWS): GWAS and fine-mapping pipelines that scale, a standard way to harmonize summary statistics, and support for new multi-omics datasets. That work involved teams from Internal Medicine, Inflammation & Immunology, Statistics, and Machine Learning & Computational Sciences.
Related: UK Biobank Pharma Proteomics Project, the pQTL data behind some of the target ranking described above.