MoDPA: Inferring Modification-dependent Protein Associations from Uniformly Reprocessed Mass Spectrometry

Abstract

Post-translational modifications (PTMs) are key regulators of protein function and cellular processes; however, the overall principles of PTM co-regulation and crosstalk remain to be fully understood. Here, we present Modification-Dependent Protein Associations (MoDPA), a framework that embeds per-site PTM detection profiles using a variational autoencoder (VAE) and builds an association network by correlating these representations. Applied to 407 reprocessed public datasets, MoDPA yields a network that partitions into clusters with distinct modification compositions and enriched pathways. Three methylation-dominated clusters are enriched for ribosome and translation pathways and contain methylated ribosomal proteins, hnRNPs and translation factors, suggesting that lysine methylation of the translation apparatus is more widespread than currently appreciated. A fourth is enriched for chromatin organisation and rRNA transcription. A mixed methylation/acetylation cluster is enriched for senescence-associated heterochromatin formation, and one cluster combines citrullination of the myelin proteins MBP and GFAP with phosphorylation of cytoskeletal and neuronal signalling proteins.