Modifying inhibitor specificity for homologous enzymes by machine learning.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Drug Discovery

Background:

• Selective enzyme inhibitors are crucial for targeted therapies and biological research.
• Designing specific inhibitors, especially for homologous enzymes, faces challenges in experimental scale and specificity tuning.
• Current machine learning (ML) approaches for protein design are limited by energy calculation accuracy and predicting multi-mutation effects.

Purpose of the Study:

• To develop and validate a novel ML-based method for designing selective protease inhibitors.
• To streamline the identification of inhibitors with tailored specificity profiles for homologous enzymes.
• To apply the method to design selective inhibitors for matrix metalloproteinases (MMPs).

Main Methods:

• Leveraging high-throughput screening (HTS) data with ML models to train predictive binding affinities.
• Designing a novel N-TIMP2 variant targeting MMP-1, MMP-3, and MMP-9.
• Experimental validation of the designed variant's binding affinity and selectivity.
• Utilizing molecular modeling and energy minimization for structural insights.

Main Results:

• Successfully designed a novel N-TIMP2 variant with a distinct specificity profile across MMP-1, MMP-3, and MMP-9.
• Experimental validation confirmed a significant specificity shift and enhanced selectivity compared to wild-type N-TIMP2.
• Structural analysis provided insights into the molecular basis of the variant's improved selectivity.

Conclusions:

• The developed ML-based method effectively reduces experimental workload in inhibitor design.
• The approach facilitates the rational design of highly selective inhibitors for homologous enzyme families.
• This work advances the understanding of enzyme-inhibitor interactions and selective targeting.