Unveiling Large-Scale Kinase-Centric Protein-Protein Interactions through a Knowledge-Informed Workflow

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Structural Biology

Background:

• Protein phosphorylation is a key regulatory mechanism in cell signaling.
• Understanding kinase-substrate specificity at the atomic level is challenging due to limited structural data and current prediction model limitations.

Purpose of the Study:

• To develop a novel computational pipeline for high-resolution modeling of kinase-substrate interactions.
• To improve the prediction of substrate specificity for protein kinases.

Main Methods:

• Reformulated kinase-substrate modeling as a Bayesian inference problem.
• Integrated curated datasets and literature evidence parsed by Large Language Models (LLMs).
• Utilized a restraint-guided deep-learning model (GRASP) and molecular dynamics for structure refinement.

Main Results:

• Generated 336 new phosphorylation-site-specific structure candidates for EGFR, BRAF, and JNK1.
• Mapped kinase recognition patches and derived sequence preferences using a Virtual Position Scanning Peptide Array (V-PSPA).
• Linked predicted kinase-substrate interfaces and pathogenic mutations to clinical data and pathogenicity scores.

Conclusions:

• The developed pipeline provides high-resolution, high-throughput modeling of kinase-substrate specificity.
• This approach has broad applicability for kinase specificity studies and drug discovery.
• The findings connect molecular interactions to pathogenicity and clinical mutations.