Improved mutant function prediction via PACT: Protein Analysis and Classifier Toolkit.

Area of Science:

• Protein engineering
• Computational biology
• Bioinformatics

Background:

• Deep mutational scanning (DMS) experiments measure the sequence-function relationship for numerous mutations simultaneously.
• Analyzing DMS data requires integrating fitness metrics with sequence and structural information for comprehensive insights.

Purpose of the Study:

• To introduce the Protein Analysis and Classifier Toolkit (PACT), a Python software package designed for analyzing DMS experiments.
• To facilitate the development of reusable and sharable protocols for custom DMS experiments.
• To demonstrate PACT's utility in assessing enzyme activity-stability trade-offs using existing DMS datasets.

Main Methods:

• PACT integrates mutation fitness data with sequence and structural features.
• The toolkit supports the creation of modular and reusable protocols for various DMS library designs (single/multiple mutations).
• PACT was used to evaluate classifiers predicting protein mutant function based on DMS data.

Main Results:

• PACT efficiently evaluated classifiers for predicting protein mutant function from DMS screens.
• Classifiers incorporating sequence homology, contact number, and proline mutations showed the best performance (low false positive, high true positive rates).
• The toolkit was exemplified on datasets analyzing enzyme activity and stability trade-offs.

Conclusions:

• PACT provides a robust framework for analyzing DMS data, linking fitness to molecular features.
• The software enables efficient evaluation of predictive models for protein function.
• PACT promotes reproducible research through sharable and modular protocols.