Hypergraph factorization for multi-tissue gene expression imputation

Abstract Integrating gene expression across tissues and cell types is crucial for understanding the coordinated biological mechanisms that drive disease characterize homoeostasis. However, traditional multi-tissue integration methods either cannot handle uncollected or rely on genotype information, which often unavailable subject to privacy concerns. Here we present HYFA (hypergraph factorization), a parameter-efficient graph representation learning approach joint imputation of cell-type expression. agnostic, supports variable number collected per individual, imposes strong inductive biases leverage shared regulatory architecture genes. In performance comparison Genotype–Tissue Expression project data, achieves superior over existing methods, especially when multiple reference are available. The HYFA-imputed dataset can be used identify replicable genetic variations (expression quantitative trait loci), with substantial gains original incomplete dataset. accelerate effective scalable tissue transcriptome biorepositories.