CATHER: a novel threading algorithm with predicted contacts.

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Protein structure prediction is crucial for understanding biological function.
• Threading algorithms are effective but can be enhanced with predicted contacts.
• Advancements in deep learning enable accurate protein contact map prediction.

Purpose of the Study:

• To develop a novel contact-assisted threading algorithm for improved protein structure prediction.
• To integrate predicted contact maps with conventional sequential profiles.
• To evaluate the performance of the new algorithm against existing methods.

Main Methods:

• Developed the CATHER (Contact-Assisted THreading) algorithm.
• Utilized both sequential profiles and deep learning-based predicted contact maps.
• Conducted benchmark tests on independent datasets and CASP targets.

Main Results:

• CATHER demonstrated significant performance improvements over methods using only sequential profiles or predicted contacts.
• The method achieved a Top 10 ranking among 39 server groups in CASP13 blind tests for free modeling targets.
• Integration of predicted contacts substantially enhances threading accuracy.

Conclusions:

• Contact-assisted threading represents a promising direction for advancing protein structure prediction.
• CATHER offers a robust approach by synergistically combining predicted contacts and sequential information.
• The findings highlight the potential of deep learning-derived contact maps in computational biology.