Are knowledge-based potentials derived from protein structure sets discriminative with respect to amino acid types?

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Biophysics

Background:

• Comparing amino acid types at protein sequence positions to structural templates requires robust residue environment descriptions.
• Current methods rely on parameters like solvent accessibility, backbone conformation, and residue-residue distances.

Purpose of the Study:

• To statistically quantify the discriminative power of amino acid type-specific distributions for commonly used environment parameters.
• To assess the reliability of sequence-structure alignment methods based on existing parameters.

Main Methods:

• Utilized the Bahadur theory to estimate error probabilities in single-sequence-structure alignments.
• Analyzed various residue environment variables, including solvent-accessible surface area, backbone conformation, and pairwise distances.
• Evaluated the discriminative power of these parameters across all 20 amino acid types.

Main Results:

• Most residue environment parameters show limited discriminative power for specific amino acid types, treating several (e.g., Ala, Asp, Gln) as 'average'.
• Complex structural characteristics also fail to distinguish many amino acid types.
• Highly discriminative amino acid types are crucial for accurate gapless alignments, while nondiscriminated types introduce ambiguities when gaps are present.

Conclusions:

• Existing sequence-structure alignment techniques relying on current parameters are unreliable, particularly when gaps are involved.
• The limited discriminative power of common parameters necessitates the invention of conceptually new and more sensitive environment descriptors.
• Future research should focus on developing novel parameters to improve the accuracy of protein sequence-structure alignments.