Variational Inference for Large-Scale Genomic Data - CGSI 2022

Explore variational inference techniques for analyzing large-scale genomic data in this conference talk from the Computational Genomics Summer Institute. Delve into the challenges of genome-wide association studies and the need for scalable inference methods in the era of massive biobanks. Learn how Bayesian approaches and variational inference can be applied to integrate molecular and functional information for understanding disease mechanisms. Discover the application of these techniques to blood GWAS, including the identification of thousands of genes associated with various traits. Examine the FactorGO method, which leverages information from under-powered studies to enhance functional genomic annotations. Gain insights into cutting-edge approaches for extracting meaningful biological information from vast genomic datasets.

Intro
Genome-wide association studies
GWAS does not provide causal mechanisms
Global biobanks reflect massive scale of available genome/phenome
Large-scale genomic analyses require scalable inference
Have you heard the good word of Thomas Bayes?
Variational approaches for approximate Bayesian inference I
Integrate molecular/functional information to understand disease mechanisms
Large-scale application to blood GWAS
TWAS identifies 6,236 and 116 genes for EA and AA across 15 traits
Gene sets by MA-FOCUS are more enriched for hematopoietic categories
Integrate phenome information to understand disease mechanisms
FactorGO: Factor analysis for genetic associations
FactorGO leverages information in under-powered studies
FactorGO finds greater enrichment at functionally relevant genomic annotations