(PS)2-v2: template-based protein structure prediction server.

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Template-based modeling (TBM) relies on accurate template selection and target-template alignment.
• Identifying templates with 15-25% sequence similarity (twilight zone) remains challenging for protein structure prediction.

Purpose of the Study:

• To present the (PS)2-v2 server, an enhanced version of a previous server, designed to improve the reliability and applicability of template-based protein structure prediction.
• To address the difficulties in template identification for proteins with low sequence similarity.

Main Methods:

• Utilizes the S2A2 matrix (a 60x60 substitution matrix incorporating secondary structure propensities) and position-specific sequence profiles (PSSMs) from PSI-BLAST for remote homology detection.
• Employs multiple templates and multiple models for building and assessing protein models.
• Leverages amino acid and structural propensities for template selection and alignment.

Main Results:

• The (PS)2-v2 server demonstrated superior performance compared to sequence-profile approaches and comparable performance to structure-based methods on the Lindahl and ProSup benchmarks.
• On the CASP8 dataset, (PS)2-v2 ranked 6th out of 72 servers for template-based modeling targets.
• The server is faster than the top-ranking ab initio methods, highlighting its efficiency.

Conclusions:

• The (PS)2-v2 server, incorporating the S2A2 matrix, effectively aids in template selection and target-template alignment by integrating amino acid and structural information.
• Multiple-template and multiple-model strategies significantly enhance alignment accuracy, particularly in the twilight zone.
• The server is a valuable tool for protein structure prediction and modeling, especially for detecting homologous templates with low sequence similarity.