Integrative Analysis Methods for the dGTEx Initiative

ABSTRACT This research project aims to develop methods and tools and conduct collaborative research for the integrative analysis of data generated by the Developmental Genotype-Tissue Expression (dGTEx) initiative, non-human primate (NHP) dGTEx project, existing GTEx project, and other studies. In Aim 1, we will develop methods for mapping

expression quantitative trait loci (eQTLs) across developmental stages in multiple tissue- and cell-types. Based on our prior work, we will employ novel multi-view learning (machine learning) methods into the proposed general QTL framework for detecting various types of QTLs. Our framework estimates the latent probabilities of

QTL binary status (presence or absence), extracts common and specific low-rank patterns from multiple groups and tissues/cell-types, and incorporates the patterns in estimating the posterior probability of non-zero effect and posterior mean/standard deviation for each input statistic. These outputs can be used for further flexible

inference in detecting various types of eQTLs. The proposed QTL framework is adaptive to a variety of integrative analyses of dGTEx, NHP, GTEx and other datasets. In Aim 2, we will develop a series of multi-age-group Mendelian randomization (MR) models to identify risk genes and assess their causal effects in multiple

tissues/cell types and age groups. We will extend the models to multi-trait analysis jointly assessing the causal effects in child and adult populations, to multivariable MR analysis accounting for other molecular traits, and to multi-cell MR analysis for detecting sparse cell-level causal effects. In Aim 3, we will engage in the dGTEx

data analysis. We will work with the Steering Committee to guarantee the scientific rigor and efficiency of dGTEx analysis, and to ensure the timely dissemination of initial findings to the broader research community. The project will develop scalable and efficient software. The insights gained through the analysis of dGTEx

data will enhance the translational potential of genomic findings in medicine and healthcare, reshaping our approach to understanding and treating diseases rooted in developmental gene regulation.