Multidimensional vector space representation for convergent evolution and molecular phylogeny

Area of Science:

• Genomics and Bioinformatics
• Molecular Evolution
• Phylogenetics

Background:

• Phylogenetic reconstruction relies on molecular evolution models, but biases can lead to incorrect evolutionary trees.
• Large genomic datasets can amplify subtle model biases, impacting phylogenetic accuracy.
• Convergent evolution and unusual substitution patterns pose challenges to accurate phylogenetic inference.

Purpose of the Study:

• To develop a novel procedure for detecting and correcting biases in molecular evolution models.
• To improve the reliability of phylogenetic tree reconstruction in the presence of evolutionary complexities.
• To address limitations in current phylogenetic methods caused by unrecognized convergent evolution.

Main Methods:

• A multidimensional vector space (MVS) approach was developed to represent sequence evolution.
• Orthogonality of sequence evolution vectors within the MVS was used to detect deviations indicative of bias.
• Biases were quantified by identifying outliers in sequence spatial vectors and modifying pairwise distances.

Main Results:

• The MVS approach successfully detected systematic biases caused by convergent evolution in various datasets.
• Bias correction using the MVS method led to a consistent phylogenetic topology across different gene datasets for placental mammals.
• Convergent evolution at specific sites explained difficulties in previous phylogenetic placements, such as for the hedgehog.

Conclusions:

• The proposed MVS method effectively identifies and corrects biases in molecular evolution models.
• Accurate phylogenetic reconstruction is achievable by accounting for convergent evolution and substitution anomalies.
• This approach enhances the robustness and consistency of phylogenetic analyses, particularly with large genomic datasets.