Protein fold recognition using HMM-HMM alignment and dynamic programming.

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Biophysics

Background:

• Protein structure prediction is crucial for understanding biological functions.
• Protein fold recognition serves as a key intermediate step in predicting 3D protein structures.
• Current feature extraction methods for protein sequences have limitations in achieving high recognition accuracy.

Purpose of the Study:

• To develop an improved method for protein fold recognition.
• To enhance the accuracy of classifying novel protein sequences into their respective folds.
• To overcome limitations of existing feature extraction techniques in protein sequence analysis.

Main Methods:

• Utilized HMM-HMM alignment (HHblits) to extract profile HMM (PHMM) matrices from protein sequences.
• Computed distances between PHMM matrices using kernelized dynamic programming.
• Evaluated the method on three benchmark datasets from the Structural Classification of Proteins (SCOP).

Main Results:

• Achieved significant improvements in protein fold recognition accuracy compared to state-of-the-art methods.
• Demonstrated accuracy improvements ranging from 2.7% to 11.6% on benchmark datasets.
• The novel feature extraction and comparison method proved effective for protein classification.

Conclusions:

• The proposed method based on PHMM matrix comparison offers a substantial advancement in protein fold recognition.
• This approach contributes to more accurate prediction of three-dimensional protein structures.
• Further research can explore the application of this method to broader biological sequence analysis tasks.