Statistical Methods for Characterizing Molecular Mechanisms of Human Tissue Development and Disease

Project Summary The unique biospecimens and data from the Developmental GTEx (dGTEx) project create an exciting opportunity and need for novel methods development. In this project, we propose to develop a set of statistical methods and analytical approaches that are designed to extract insights into human and non-human primate

development from the multi-modal and multi-tissue data from dGTEx post-mortem donors. Analysis of the dGTEx data requires statistical models that can take advantage of the rich structure and diversity of the data across ages and modalities, while addressing some of the inherent challenges. Models explicitly informed by

age can capture developmental trajectories of gene regulation, genetic effects, and tissue structure. The range of data modalities also creates an opportunity for novel methods to capture additional effects and improved resolution at the cell-type, isoform, and structural levels. However, these data are also inherently complex,

representing a mixture of cell types along with biological and technical noise. The ambitious study design of dGTEx also comes with challenges of donor recruitment that limit sample size. The methods proposed here are designed to leverage the benefits of temporal multi-modal data, while addressing data complexity and

limited sample size. In our first aim, we will analyze transcriptome variation across the human lifespan, with improved transcript annotation and new methods to characterize how gene expression, alternative splicing and cell type composition change during development and how they contribute in driving phenotypic change.

Secondly, we will use multi-modal data from GTEx to capture changes in gene regulatory networks during development. From the dGTEx histology images and spatial transcriptomics data we will model developmental trajectories of tissue structures, and describe their molecular characteristics as well as role in disease. In our

third aim we will map and characterize genetic regulatory variation in dGTEx and apply predictive models for improved predictions of regulatory variants in pediatric tissues. In addition to empowering biological discovery, this work has the potential to uncover disease risk factors and mechanisms that originate or manifest during

early life.