Remote protein homology detection using recurrence quantification analysis and amino acid physicochemical properties

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Biology

Background:

• Remote homology detection is crucial for understanding protein evolution and function, especially for proteins with low sequence similarity.
• Current methods often rely on sequence alignment or profiles, with limited use of amino acid physicochemical properties.
• Support Vector Machine (SVM) algorithms are state-of-the-art but can be improved with novel feature representations.

Purpose of the Study:

• To develop a novel remote homology detection method (SVM-RQA) that incorporates amino acid physicochemical properties and Recurrence Quantification Analysis (RQA).
• To evaluate the performance of SVM-RQA against existing methods on a standard protein dataset.
• To explore the potential of physicochemical properties for biological interpretation in protein family classification.

Main Methods:

• Representing protein primary sequences using selected amino acid indices based on physicochemical properties.
• Measuring protein similarity using Recurrence Quantification Analysis (RQA) metrics.
• Employing an SVM classifier on the RQA-derived feature space.

Main Results:

• The SVM-RQA method achieved classification accuracy comparable to state-of-the-art SVM kernels without using sequence alignment or profile information.
• An optimization scheme successfully identified optimal amino acid indices for characterizing protein families.
• The approach demonstrates the utility of physicochemical properties for remote homology detection.

Conclusions:

• The SVM-RQA method offers a promising alternative for remote homology detection, leveraging intrinsic protein properties.
• Incorporating physicochemical properties provides a potentially richer biological insight into protein family classification.
• Future work could enhance accuracy by combining alignment-based and property-based features.