Large-Scale Structure-Based Prediction and Identification of Novel Protease Substrates Using Computational Protein...

Area of Science:

• Biochemistry
• Computational Biology
• Enzymology

Background:

• Protease enzymes play vital roles in biological processes, necessitating accurate substrate specificity prediction.
• Current in silico methods for protease specificity prediction are limited by reliance on existing experimental data and sequence patterns.

Purpose of the Study:

• To develop a general, de novo approach for predicting peptidase substrates using protein structure modeling and biophysical evaluation.
• To improve the accuracy and discriminatory power of protease specificity prediction compared to existing sequence-based methods.

Main Methods:

• Constructed atomic resolution models of enzyme-substrate complexes for five model proteases across four mechanistic classes.
• Developed a discriminatory scoring function using Rosetta and AMBER's MMPBSA, ranking substrates by interaction energy with modeled active site conformations.
• Combined sequence and energetic patterns with machine learning for enhanced classification performance.

Main Results:

• Energetic patterns from simulations robustly ranked and classified known cleaved and uncleaved peptides.
• Structural-energetic patterns demonstrated superior discriminatory power over purely sequence-based inference.
• Experimental validation using hepatitis C virus NS3/4 protease confirmed the model's predictive capability for novel substrate motifs.

Conclusions:

• The developed structure-based approach provides a generalizable and accurate method for predicting protease substrate specificity.
• This method complements experimental techniques and offers physical insights into enzyme-substrate interactions.
• The approach has potential applications in designing custom proteases for biotechnological and therapeutic purposes.