Human variation in population-wide gene expression data predicts gene function and phenotype

Population-scale multi-layered datasets assemble extensive experimental data of different types on single healthy individuals in large cohorts, capturing genetic variation and environmental factors influencing gene expression with no clinical evidence of pathology. Variance of gene expression can be exploited to set up a conditional quasi loss- and gain-of-function “in population” experiment if expression values for the gene of interest (GOI) are available. We describe here a novel approach, called huva (human variation), that takes advantage of population-scale multi-layered data to infer gene function and relationships between phenotypes and gene expression. Within a reference dataset, huva derives two experimental groups, i.e. individuals with LOW or HIGH expression of the GOI, enabling the subsequent comparison of their transcriptional profile and functional parameters. We demonstrate that this approach robustly and efficiently identifies the phenotypic relevance of a GOI, allows the stratification of genes according to shared biological functions, and we further generalized this concept to almost 16,000 genes in the human blood transcriptome. Additionally, we describe how huva predicts the phenotype of naturally occurring activating mutations in humans. Here, huva predicts monocytes rather than lymphocytes to be the major cell type in the pathophysiology of STAT1 activating mutations, evidence which was validated in a cohort of clinically characterized patients. This repository contains the huva package (v 0.1.4) used in the original manuscript, Bonaguro et al. iScience 2022, together with the R enviroment of the analysis shown in the manuscript (https://github.com/lorenzobonaguro/huva_reproducibility