Penn-disam: data integration and statistical analysis methods for dgtex data

The University of Pennsylvania received a $1,529,885 Cooperative Agreement from the National Human Genome Research Institute (NHGRI) under the Human Genome Research program (CFDA 93.172) to develop novel statistical and computational methods for analyzing developmental gene expression data from the DGTEX (Developmental Genotype-Tissue Expression) and NHP DGTEX projects. The overarching goal is to create tools for effective analysis of age-dependent gene expression functions across different tissues and between human and non-human primate models. Key components include nonparametric regression methods, data visualization, statistical inference, and techniques to align gene expression trajectories between species. The Children's Hospital of Philadelphia (CHOP) is a sub-awardee providing data management and bioinformatics expertise to support the project. The work is expected to run from September 2024 through August 2027.

Abstract

Our health and wellbeing prenatally and during the earlier years of life affect all future health and disease risks. This time period is the most sensitive for a child's developing brain and various other tissues of body. Studies of developmental biology have demonstrated that gene expression patterns are not only tissue-specific and cell type-specific, but also age regulated and controlled through coordinated action of complex tissue- and cell-type specific networks and pathways. The NIH developmental genotype-tissue expression (DGTEX) project aims to study gene expression patterns at a tissue-level over several early developmental windows and to char- acterize transcriptional profiles during human development. Complementary to the human DGTEX effort, the non-human primate (NHP) DGTEX aims to study gene expression patterns in multiple reference tissues across developmental stages in NHP model species and compare them to human gene expression patterns. The over- arching goal of PENN data integration and statistical analysis methods (PENN-DISAM) project is to develop novel statistical and computational methods specifically for effective analysis of human DGTEX and NHP DGTEX data, including novel methods for analyzing a very large set of correlated regression functions that characterize the age-dependent gene expression functions across different tissues and between human and primates. The DG- TEX and NHP DGTEX data allow US to estimate the tissue-cell specific age-dependent gene expression functions by leveraging the gene expression data measured over different ages in postnatal, early childhood, pre-pubertal and post-pubertal developmental windows. We will particularly develop nonparametric B-spline regression to estimate the age-dependent gene expression regression functions and summarize the data as the matrix of functions. We will develop methods for data visualization and for statistical inference, including methods for iden- tifying genetic variants that are associated with different gene expression distribution functions in each of the tissues. We will also develop novel statistical warping methods to align gene expression trajectories between human and primates, which allow US to identify genes under strong stabilizing selection, to quantify the inter- species divergence in gene expression in each tissue and in each developmental stage, and to compare the difference of age-dependent weighted CO-expression networks between human and primates. PENN-DISAM will work closely with the DGTEX consortium to develop, implement, test and apply these methods and the software tools to DGTEX data. Finally, we will make all the software available via anvil and github.