A Data- and Model-Driven Analysis Reveals the Multi-omic Landscape of Ageing

Altered expression of a number of genes has been correlated to biological ageing in humans. The biological age predicted from gene expression levels is known as transcriptomic age. This differs from chronological age which is measured as the time that an individual has lived since their date of birth. Transcriptomic age can be older or younger than an individual’s chronological age. At present, studies have focused on using transcriptomic data to predict transcriptomic age. However, this approach largely does not consider the effect that genes have on the metabolic network and therefore on the observable cellular phenotype. This research takes the current understanding of transcriptomic ageing a step further by generating and investigating genome-scale metabolic models of ageing, using machine learning methods and a multi-omic approach based on constraint-based modelling. We combine these models with a transcriptomic age predictor and gene expression data from CD4 T-Cells from human peripheral blood mononuclear cells in healthy individuals. We show that metabolic models augmented with transcriptomics data of ageing can generate greater metabolic insights into the differences between chronological and transcriptomic age. Compared to standard transcriptomic-only approaches, our method provides a more comprehensive analysis of transcriptomic ageing and paves the way for a multiomic understanding of ageing mechanisms in human cells.