Machine Learning for Biobank-Scale Genomic Data - CGSI 2022

Explore machine learning techniques for analyzing Biobank-scale genomic data in this comprehensive conference talk from the Computational Genomics Summer Institute. Delve into key inference problems and the genetic architecture of complex traits, focusing on variance components models and efficient estimation methods. Learn about the Randomized HE-regression (RHE) approach for accurate and scalable analysis of large-scale genomic data. Examine insights gained from applying RHE to the UK Biobank, including dominance deviation effects, gene-environment interactions, and gene-gene interactions. Discover how Random Fourier Features (RFF) can be used to address higher-order effects and tackle the challenge of missing data in Biobanks. Gain valuable knowledge on cutting-edge machine learning applications in genomics, supported by related research papers on variance components analysis, dominance deviation effects, and population structure inference in biobank-scale data.

Intro
Machine learning for genomic data
Growth of Biobanks
Key inference problems
Genetic architecture of complex traits
Variance components model
Estimating variance components
Alternate estimator Method of Moments (HE-regression)
Randomized HE-regression (RHE) Work with a "sketch" of the genotype
RHE is accurate and scalable
Insights from applying RHE to UK Biobank
Dominance deviation effects
Dominance deviance effects
Gene-environment interactions (GxE)
Gene-gene interactions (GxG)
Beyond pair-wise effects
Random Fourier Features (RFF)
Missing data in Biobanks