Finally considering the utility of PGSxE, we introduced the notion of proportion needed to benefit (PNB) as the cumulative number needed to treat across PGS thresholds in high vs low-risk exposures & show that it is typically halved between 70th–80th PGS percentile. [8/9]

The predominant mechanism for PGS×E interactions is shown to be amplification of genetic effects in the presence of adverse exposures such as low polyunsaturated fatty acids, mediators of obesity, and social determinants of ill health. [7/9]

Predictive accuracy is significantly improved in the high-risk (adverse) exposures and by including interaction terms with effects as large as those documented for low transferability of PGS across ancestries. [6/9]

Across all disease-exposures, we find evidence of pervasive PGSxE interactions influencing common disease risk. Eg. for incident CAD, key exposures exhibiting multiple interactions are: sex, weekly beer intake, smoking and omega-6 fatty acids. [4/9]

For example, for coronary artery disease (CAD): Low levels of omega-6 fatty acids and past tobacco smoking interact with PGS-CAD to exacerbate incident CAD risk. [3/9]

The impact of PGS on the disease is highly context-specific. We quantify polygenic score-by-exposure (PGSxE) interactions for seven common diseases and pairs of 75 exposures. [2/9]

This year has been special - my doctoral graduation ceremony with my advisor and mentor, Greg Gibson (@genomestake ) and my family 🙂

Very excited to have presented my Reviewer's Choice poster at #ASHG22 on predicted TRS supporting the evidence of canalization of polygenic risk for common diseases and traits in the UK Biobank - PB1592

Honored to be selected as the Postdoctoral Semifinalist for 2022 Charles J. Epstein Awards for Excellence in Human Genetics Research for #ASHGef="/hashtag/ASHG22" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link">#ASHG22! Plus, our abstract has been selected as the Reviewer's Choice abstract 😀 #ASHG @GeneticsSociety...

Lifestyle related exposures show decanalization for BMI but canalization for WHR, reflecting different evolutionary pressures on the architectures of weight-related traits. Could be explained by recent human behaviors driving BMI vs stabilizing selection for metabolism for WHR.

For continuous traits: Decanalization for BMI wrt Townsend deprivation index

All exposures vs college attainment

(De)canalization is defined based on the observed vs expected deviations at the extremes (Delta) and the departure from null expectation above or below a certain threshold.
All exposures vs CAD

For binary traits, we compared the observed prevalence with the expected prevalence under null; Expected prevalence computed using liability threshold model assuming environmental effects to be additive (no PGSxE).
Past tobacco smoking & coronary artery disease risk

C. Walk pace vs obesity risk: As the PGS increase, genetic effects appear to be larger in the poor environment (slow walk pace), implying genetic variance is greater at the extremes, leading to higher impact on disease risk - decanalization.

Using 10 complex traits and 151 environmental exposures, we compare the prevalence-risk curves wrt exposures to ask the questions 1.For which trait-exposure combination does PGSxE exist? If it exists, can we develop a modeling framework to assess decanalization vs canalization?

Reading @kph3k's wonderfully written "The Genetic Lottery", while beginning the journey to write my PhD thesis🙂
Any tips and suggestions as I embark on this journey are welcome!

Here we show the prediction of Colectomy in Ulcerative Colitis patients using Transcriptional Risk Scores (TRS) and also TWAS-based Predicted Polygenic Transcriptional Risk Scores (PP-TRS), both replicated in independent cohorts.

All set for the Summer Institute of Statistical Genetics at University of Washington, Seattle!
Module 1: Bayesian Statistics for Genetics by Jon Wakefield and Ken Rice @UWBiostat #SISG #statgen