The limits of protein secondary structure prediction accuracy from multiple sequence alignment

Area of Science:

• Protein structure and bioinformatics
• Computational biology and structural bioinformatics

Background:

• Assessing secondary structure prediction accuracy is crucial for understanding protein folding and function.
• Current methods often face challenges in evaluating consensus predictions against individual protein structures.

Purpose of the Study:

• To determine expected residue-by-residue accuracies for secondary structure prediction using multiple protein sequence alignments.
• To establish a benchmark for evaluating the quality of consensus secondary structure predictions.

Main Methods:

• Analysis of known protein structural families.
• Comparison of consensus predictions with individual homologous protein structures.

Main Results:

• Significant variation in secondary structure exists among homologous proteins.
• 100% agreement between consensus predictions and individual family members is unlikely.
• The study defines the expected range of agreement for perfect secondary structure predictions.

Conclusions:

• The established ranges overcome limitations in assessing consensus prediction accuracy.
• Recent predictions for annexins, SH2, and SH3 domains align with expected perfect prediction accuracies.