Tarmo Nurmi, Natasa Przulj, Noel Malod, Stevan Milinkovic, Branislava Jankovic and Aleksandr Matsun*
Mohamed Bin Zayed University of Artificial Intelligence
aleksandr.matsun [at] mbzuai.ac.ae
Abstract
Aging is a complex biological process that affects almost all systems of an organism and is known to be a major driver of numerous diseases. Despite a lot of research on aging in general, few studies look at it from a pan-tissue perspective, thus availability of systemic anti-aging interventions remains limited. In this work we present a novel approach that integrates age- and tissue-specific bulk RNA sequencing data with prior knowledge in the form of bulk proteomics and metabolomics data using non-negative matrix tri-factorization. By analyzing the dynamics of genes’ learned embeddings across all the age brackets, our model curates an atlas of tissue-specific aging-associated genes, which we use afterwards to perform a systematic selection of pan-tissue aging-related genes. Applying this framework to the multi-omics data that includes gene expressions in 31 tissues from GTEx database, we derive and validate a set of 21 genes that undergo significant changes with age in the vast majority of tissues. Finally, out of those 21 genes we consider 7 to be promising therapeutic targets, due to their involvement in such molecular processes as inflammation, metabolism and stress response, and propose potential intervention strategies that use drug repurposing and de novo drug generation. Our method provides a scalable approach for integrative multi-omics analysis of aging that can be further extended to additional datasets.
Keywords: Aging, Multi-omics, Pan-tissue, Integration