Deep learning model of post-translational modification regulating liquid-liquid phase separation.

Liquid-liquid phase separation (LLPS) drives the formation of various membraneless organelles, which are crucial for biological processes and disease development. Despite the significant regulatory effects on LLPS of protein post-translational modifications (PTMs), specific data resource and predictor are still lacking. First, we constructed a well-curated database of PTM regulation on liquid-liquid Phase Separation (PTMPhaSe) ( https://ptmphase.sjtu.edu.cn ) that contains manually curated complete experimental evidence. Second, we developed graph neural network-based deep learning model (named PhosLLPS) to predict functional phosphorylation sites regulating LLPS, which achieved better identification performance (AUC = 0.9116) than four baseline models and the existing FuncPhos-SEQ method. Meanwhile, human proteome-scale predictions for functional phosphorylation sites were performed with PhosLLPS. PhosLLPS is now freely available in web server ( https://ptmphase.sjtu.edu.cn/Predictor ). By bridging the gap between PTM regulation and LLPS, these resources could contribute to a better understanding of the molecular function of LLPS and facilitate further drug development for LLPS-related diseases.