Combining distance matrices on identical taxon sets for multi-gene analysis with singular value decomposition.

Area of Science:

• * Evolutionary Biology
• * Bioinformatics
• * Computational Biology

Background:

• * Phylogenetic analysis often relies on single genes, which can lead to incomplete or conflicting evolutionary signals.
• * Combining data from multiple genes is crucial for robust phylogenetic inference but presents methodological challenges.

Purpose of the Study:

• * To develop a simple and effective method for combining distance matrices from multiple genes.
• * To derive a single representative distance matrix for constructing a combined-gene phylogenetic tree.
• * To assess the method's performance and identify conflicting signals within datasets.

Main Methods:

• * Application of singular value decomposition (SVD) to distance matrices from multiple genes.
• * Extraction of the greatest common signal using the first right eigenvector of the SVD.
• * Derivation of a representative tree from the weighted average of gene distance matrices.
• * Uncertainty estimation using bootstrap methods and comparative analysis with existing methods (SDM, SDM*, ACS97).

Main Results:

• * The SVD-based method successfully generates a representative phylogenetic tree from multiple genes.
• * Uncertainty estimates for the derived trees were found to be small across tested datasets.
• * The method's performance was comparable to other distance-based approaches in consensus settings.
• * Subsequent eigenvalues and eigenvectors can reveal conflicting signals and identify influential or outlier genes.

Conclusions:

• * The proposed SVD-based method offers a powerful tool for integrated phylogenetic analysis using multi-gene data.
• * This approach enhances the reliability of phylogenetic trees by effectively averaging signals and quantifying uncertainty.
• * The method provides additional insights into data conflicts and gene influence, aiding in critical evaluation of phylogenetic results.