Identification of substrates for Ser/Thr kinases using residue-based statistical pair potentials

Area of Science:

• Computational biology
• Biochemistry
• Bioinformatics

Background:

• In silico methods are crucial for identifying kinase substrates and understanding cell signaling.
• Current prediction methods rely on experimental data, limiting their scope to known kinase families.
• A novel approach is needed that bypasses the requirement for experimental substrate data.

Purpose of the Study:

• To develop a novel, multi-scale, structure-based computational approach for predicting kinase substrates.
• To create a method that does not require prior experimental substrate data for training.
• To enhance the accuracy and applicability of in silico kinase substrate prediction.

Main Methods:

• Utilized residue-based statistical pair potentials for scoring substrate peptide-kinase binding energy.
• Developed a multi-scale structure-based computational strategy.
• Employed all-atom force fields and MM/PBSA for refining high-scoring candidate substrates.

Main Results:

• The novel method was benchmarked on the Phospho.ELM dataset.
• Demonstrated superior performance compared to existing structure-based prediction methods.
• Achieved prediction accuracy comparable to sequence-based methods.
• Showcased improved substrate ranking through all-atom force field and MM/PBSA modeling.

Conclusions:

• The developed structure-based method offers a powerful alternative for kinase substrate prediction.
• This approach expands the applicability of in silico methods to kinases lacking experimental substrate data.
• The combination of statistical potentials and advanced modeling techniques enhances prediction accuracy.