Deciphering the functional landscape of phosphosites with deep neural network

• DOI: 10.1016/j.celrep.2023.113048
• Publication Date: 2023-09-01T17:20:01Z
• Authors: Zhongjie Liang; Tonghai Liu; Qi Li; Guangyu Zhang; Bei Zhang; Xikun Du; Jingqiu Liu; Zhifeng Chen; Hong Ding; Guang Hu; Hao Lin; Fei Zhu; Cheng Luo
• Keywords: Human proteome project

Current biochemical approaches have only identified the most well-characterized kinases for a tiny fraction of phosphoproteome, and functional assignments phosphosites are almost negligible. Herein, we analyze substrate preference catalyzed by specific kinase present novel integrated deep neural network model named FuncPhos-SEQ assignment human proteome-level phosphosites. incorporates phosphosite motif information from protein sequence using multiple convolutional (CNN) channels features protein-protein interactions (PPIs) embedding (DNN) channels. These concatenated jointly fed into heterogeneous feature to prioritize Combined with series in vitro cellular assays, confirm that NADK-S48/50 phosphorylation could activate its enzymatic activity. In addition, ERK1/2 discovered as primary responsible phosphorylation. Moreover, is developed an online server.