Physics-Based Energy Functions for Computational Protein Design

Area of Science:

• Biochemistry
• Computational Biology
• Protein Engineering

Background:

• Computational protein design (CPD) aims to engineer novel proteins with specific functions or structures using computational approaches.
• The inverse folding problem, predicting protein sequences from a given backbone, is a key area within CPD with a 40-year history of methodological development and experimental success.
• Scoring functions are crucial for evaluating and comparing potential protein sequences and conformations.

Purpose of the Study:

• To review computational protein design works that utilize physics-based energy functions.
• To discuss the advantages, interests, and future perspectives of physics-based methods in CPD.
• To compare physics-based approaches with other methods, including statistical, empirical, and deep learning-based techniques.

Main Methods:

• Review of literature focusing on computational protein design methodologies.
• Analysis of scoring functions, categorizing them into statistical, empirical, and physics-based approaches.
• Discussion of the strengths and weaknesses of different CPD methods, particularly highlighting physics-based energy functions.

Main Results:

• Physics-based energy functions provide greater explanatory power and do not rely on training datasets, unlike deep learning methods.
• Deep learning approaches have shown undeniable improvements in prediction performance for CPD.
• Physics-based methods have a long history of success in designing new protein folds and enzymatic functions.

Conclusions:

• Physics-based energy functions remain valuable in computational protein design due to their interpretability and data independence.
• While deep learning offers performance gains, physics-based methods provide deeper insights into protein behavior.
• Future research in CPD should consider the complementary strengths of both physics-based and data-driven approaches.