Computational prediction of protein folding rate using structural parameters and network centrality measures

Area of Science:

• Computational biology
• Biophysics
• Structural biology

Background:

• Protein folding is a complex process determining a protein's 3D structure.
• Existing methods struggle to accurately predict protein folding rates (ln(k_f)) for all protein types.
• Machine learning (ML) models show promise but lack mechanistic explanations.

Purpose of the Study:

• To evaluate ML algorithms for predicting ln(k_f).
• To assess the utility of structural parameters and network centrality measures in folding rate prediction.
• To identify a robust model for predicting protein folding rates.

Main Methods:

• Evaluated ten ML algorithms using eight structural parameters and five network centrality measures.
• Utilized newly constructed datasets for training and validation.
• Compared the predictive performance of different algorithms and feature combinations.

Main Results:

• Support vector machine (SVM) demonstrated superior performance in predicting ln(k_f) across two-state (TS), non-two-state (NTS), and combined datasets.
• SVM achieved mean absolute differences of 1.856 (TS), 1.55 (NTS), and 1.745 (combined).
• Combining structural parameters and network centrality measures enhanced prediction accuracy compared to individual features.

Conclusions:

• SVM is a highly effective tool for predicting protein folding rates.
• Integrated analysis of structural parameters and network centrality provides a more comprehensive understanding of protein folding.
• This approach offers improved predictive power for diverse protein folding dynamics.