Integrating 3D genomic and epigenomic data to enhance target gene discovery and drug repurposing in transcriptome-wide association studies

Abstract

Transcriptome-wide association studies (TWAS) are popular approaches to test for association between imputed gene expression levels and traits of interest. Here, we propose an integrative method PUMICE (Prediction Using Models Informed by Chromatin conformations and Epigenomics) to integrate 3D genomic and epigenomic data with expression quantitative trait loci (eQTL) to more accurately predict gene expressions. PUMICE helps define and prioritize regions that harbor cis-regulatory variants, which outperforms competing methods. We further describe an extension to our method PUMICE +, which jointly combines TWAS results from single- and multi-tissue models. Across 79 traits, PUMICE + identifies 22% more independent novel genes and increases median chi-square statistics values at known loci by 35% compared to the second-best method, as well as achieves the narrowest credible interval size. Lastly, we perform computational drug repurposing and confirm that PUMICE + outperforms other TWAS methods.

Multi-ancestry and multi-trait genome-wide association meta-analyses inform clinical risk prediction for systemic lupus erythematosus

“…Specifically, we first derived gene expression prediction models using two reference datasets from disease-relevant tissues, i.e., the Genetic European Variation in Disease 39 (GEUVADIS; lymphoblastoid cell line (LCL); n = 358) and Depression Gene Network 40 (DGN; whole blood; n = 873) datasets. We used a new method PUMICE that integrates 3D genome and epigenetic information to improve prediction accuracy 41 . To assess the accuracy of prediction models, we calculate Spearman’s correlation coefficients between measured and predicted gene expression using nested cross-validation as described in Khunsriraksakul et al 41 and assess whether Spearman’s correlation is significantly different from zero 41 .…”

“…We used a new method PUMICE that integrates 3D genome and epigenetic information to improve prediction accuracy 41 . To assess the accuracy of prediction models, we calculate Spearman’s correlation coefficients between measured and predicted gene expression using nested cross-validation as described in Khunsriraksakul et al 41 and assess whether Spearman’s correlation is significantly different from zero 41 . Using PUMICE, we obtained 7028 and 9260 significant genes with Spearman’s correlation coefficients >0.1 and the corresponding P values <0.05 from GEUVADIS and DGN, respectively.…”

“…Gene expression prediction models were constructed via PUMICE 41 . Briefly, PUMICE utilizes epigenetic information to prioritize essential genetic variants that carry important functional roles.…”

“…As the genes may be correlated due to linkage disequilibrium, we calculate the correlation between squared TWAS Z score statistics via bootstrap. For the Z-score statistic from two genes 1 and 2, Z 1 and Z 2 , the correlation between the TWAS Z-statistic can be calculated from LD panel as described in Khunsriraksakul et al 41 , which we denote as ρ . To calculate the correlation between squared Z statistics, we simulate bivariate normal distributed statistics and , b = 1,…, B , where We then calculate the correlation between squared TWAS statistics as We denote the correlation matrix for the squared Z statistics of G genes as Ρ .…”

Chromosome conformation capture approaches to investigate 3D genome architecture in Ankylosing Spondylitis

“…These structures can be impacted by disease-associated SNPs, as reported extensively ( Gorkin et al, 2019 ; Anania and Lupianez, 2020 ; Tsuchiya et al, 2021 ). Recent studies in polygenic disorders ( Girdhar et al, 2022 ) show that using 3D genome architecture investigation has the utility to clarify the role of disease associated SNPs and to link them to specific genes to understand the phenotype and account for biological function ( Khunsriraksakul et al, 2022 ; Zhao et al, 2022 ).…”