Swfoldrate: predicting protein folding rates from amino acid sequence with sliding window method

Area of Science:

• Biochemistry
• Computational Biology
• Structural Biology

Background:

• Protein folding is crucial for biological function, determining a protein's active conformation from its denatured state.
• Predicting protein folding rates is challenging, with existing methods often requiring tertiary structure information.
• The relationship between amino acid sequences and protein folding rates is a key area of research.

Purpose of the Study:

• To develop a novel method for predicting protein folding rates directly from amino acid sequences.
• To eliminate the need for tertiary structure information in folding rate prediction.
• To identify key features influencing protein folding rates.

Main Methods:

• Utilized long-range and short-range protein contacts to derive an extended pseudo amino acid composition via a sliding window method.
• Employed optimal feature selection combining forward and sequential backward selection.
• Developed a nonlinear support vector machine (SVM) regression model incorporating physicochemical and statistical features.

Main Results:

• Achieved accurate prediction of protein folding rates solely from amino acid sequences.
• Demonstrated the method's effectiveness on a large dataset using jackknife cross-validation.
• Obtained a high correlation coefficient (0.9313) and low standard error (2.2692) between predicted and observed folding rates.

Conclusions:

• The developed method accurately predicts protein folding rates from sequence data alone.
• This approach offers a valuable tool for structural biology and drug discovery.
• A freely available prediction server (http://www.jci-bioinfo.cn/swfrate/input.jsp) has been established.