SSPSPredictor(Sequence and Structure based PSPs Predictor) takes sequence and structure as input. It utilizes the sequence embedding information from the large language model ESM-2, as well as a graph neural network to incorporate protein structure information based on the results of Alphafold 2.
SSPSPredictor is not just a binary classifier for phase-separating proteins. It goes beyond traditional prediction by providing:
1.Amino Acid-Level Analysis: Utilizing Attention scores, SSPSPredictor identifies key amino acids critical for driving phase separation.
2.Phase Separation Propensity: Quantitative insights into a protein's ability to undergo phase separation.