Phosphorylation modifications regulate protein-protein interactions (PPIs) to form complex signaling networks within cells. Abnormal phosphorylation states can alter PPIs and perturb normal cellular processes, leading to serious diseases such as cancer and neurodegenerative disorders. Herein, a novel integrated deep neural network model named PhosPPI-SEQ, in which the pre-trained transformer is integrated with cross-attention, is proposed for predicting the functional phosphosites with PPI regulation at the human proteome level. By leveraging sequence features extracted from pre-trained transformer based protein language model (pLM), cross attention was utilized to integrate the sequence features of phosphosite motifs with interacting proteins, allowing PhosPPI-SEQ to capture the complex interactions between the functional phosphosites and interacting proteins.

PhosPPI-SEQ requires only the raw protein sequences of the concerned PPI, and concerned phosphosites as input, avoiding the problem of losing information for a large amount of phosphosites due to the limited number of protein crystal structures. It is computationally efficient and has a broader scope of application for biologists.