Functional Enrichment of Lipidomics Data to Characterize the Metabolism of Diabetes Mellitus and Subclinical Carotid Atherosclerosis

Pathway enrichment analysis aims to uncover global functional patterns in omics data, including metabolomics. However, its application to lipidomics, the comprehensive study of lipids in biological systems, is hindered by the limited availability of lipid-specific pathways in current databases. Tools such as LINEX2and LION/web offer solutions for functional enrichment of lipids. Nevertheless, more information can still be extracted from large databases like LIPID MAPS. In this work, we present a novel functional enrichment algorithm based on a hierarchical graph constructed through unsupervised clustering of molecular embeddings. These embeddings were generated by applying a SMILES (Simplified Molecular Input Line Entry System) language model to all lipids in LIPID MAPS. We apply this algorithm to five comparisons involving type 1 and type 2 diabetes, as well as subclinical carotid atherosclerosis in type 2 diabetes. Additionally, we perform enrichment using LINEX2on the same datasets and compare the outcomes to highlight complementary insights.