Scoredist: a simple and robust protein sequence distance estimator

Area of Science:

• Bioinformatics
• Computational Biology
• Evolutionary Biology

Background:

• Distance-based methods are widely used for phylogenetic tree reconstruction due to their efficiency and broad applicability.
• Accurate estimation of evolutionary distances from sequence alignments is crucial, often requiring corrections for multiple substitutions.
• Existing accurate protein distance estimators, such as optimal matrix methods, are computationally intensive and sensitive to the chosen evolutionary model.

Purpose of the Study:

• To develop a novel, fast, and robust protein sequence distance estimator.
• To evaluate the performance of the proposed estimator against existing methods across different evolutionary models.
• To provide a distance estimation method that is less dependent on specific evolutionary models.

Main Methods:

• Proposed Scoredist, a correction-based protein sequence estimator utilizing logarithmic correction based on BLOSUM62 alignment scores.
• Evaluated Scoredist and optimal matrix methods using simulated alignments with known evolutionary distances (0.01-2 substitutions per site).
• Tested methods across three evolutionary models (Dayhoff, Jones-Taylor-Thornton, Muller-Vingron) for training and testing, and Whelan and Goldman for testing.

Main Results:

• Scoredist demonstrated accuracy comparable to optimal matrix methods but with significantly greater robustness.
• Scoredist exhibited superior accuracy when trained on one evolutionary model and tested on a different one.
• Traditional Jukes-Cantor and Kimura correction methods were found to be substantially less accurate than Scoredist.

Conclusions:

• Scoredist is a fast, accurate, and robust protein distance estimator.
• Its model-independent nature enhances its reliability in phylogenetic analyses.
• Scoredist has been integrated into the Belvu alignment viewer for practical application.