Metabolite integration pipeline for the improvement of human metabolic models

Abstract

Genome-scale metabolic models (GSMMs) of human cells are predictive tools with great potential for revealing important aspects of cell physiology, disease as well as for the diagnosis and treatment of diseases caused by the deregulation of metabolism. In the past decade, there have been notable efforts to reconstruct models of human metabolism, with five generic GSMMs currently available. Maintaining references to biological databases is important to allow seamless integration of models themselves and with experimental data. Still, the incorporation of external identifiers is often missed in the model reconstruction process. In this work, we review the most relevant human GSMMs, analyze the presence of external database identifiers, extract available metabolite annotation and identifiers and create an internal database of metabolites. Using a graph-based system loaded with information from the most relevant omics data repositories, we attempt to cluster similar metabolites through database cross-referencing. With this approach, we have successfully enriched the metabolite annotation of several older GSMMs and identified common entities that could be leveraged in the future towards the creation of a unified consensus model of human metabolism.