Computational Problems in Multi‑tissue Models of Health and Disease

Abstract

A modern development at the interface of computer science
and systems biology is being fostered by high-dimensional molecular
data emerging on multiple tissues of the same individual collected
across large groups of healthy/diseased individuals. We review computational
and statistical problems that arise in analyzing such multi-tissue
genomic datasets, specifically problems posing new challenges compared
to their single-tissue counterparts, such as ones related to missing
data imputation, statistical learning of high-dimensional network models
capturing gene–gene correlations within/across tissues, and graph algorithms
to identify genes clustering across many tissue networks. A recurring
research theme is the potential to integrate or pool information from
across tissues to enhance power of detecting signals shared across tissues
while also accounting for tissue-specific differences. We show how
methods harnessing this integrative potential to address multi-tissue
problems ranging from correlation/causal network inference to graph
algorithms are ushering in an era of integrated, whole-system modeling
of life processes.