An experimental phylogeny to benchmark ancestral sequence reconstruction

Area of Science:

• Molecular Evolution
• Biophysics
• Bioinformatics

Background:

• Ancestral sequence reconstruction (ASR) is a valuable tool for understanding molecular evolution.
• A key limitation of ASR is the difficulty in validating its algorithms experimentally.
• This study addresses the need for biological validation of ASR methods.

Purpose of the Study:

• To experimentally validate ancestral sequence reconstruction (ASR) algorithms.
• To compare the phenotypic accuracy of different ASR methods.
• To assess the impact of extant sequence sampling on ASR.

Main Methods:

• An experimental phylogeny was constructed using a red fluorescent protein gene.
• 19 extant sequences (leaves) and 17 ancestral nodes were generated.
• ASR analyses were performed using various algorithms, including Bayesian methods and maximum parsimony.
• Inferred ancestral sequences and phenotypes were benchmarked against known experimental data.

Main Results:

• All tested ASR algorithms demonstrated high accuracy in inferring ancestral sequences.
• Significant variation was observed in the phenotypes encoded by incorrectly inferred sequences.
• Bayesian methods incorporating rate variation significantly outperformed maximum parsimony in phenotypic accuracy.
• Subsampling extant sequences had a minimal impact on the accuracy of ancestral sequence inference.

Conclusions:

• Experimental validation confirms the high accuracy of ASR algorithms for sequence inference.
• Phenotypic accuracy is a critical metric for evaluating ASR, with Bayesian methods showing an advantage.
• The study provides a robust biological framework for benchmarking ASR tools.