kdetrees: Non-parametric estimation of phylogenetic tree distributions

Area of Science:

• Phylogenetics
• Computational Biology
• Bioinformatics

Background:

• Most gene histories within a clade follow a common evolutionary process.
• Deviations from this common process, such as horizontal gene transfers or gene duplication, result in 'outlying' gene trees.
• Identifying these outlying gene trees is important for understanding complex evolutionary histories.

Purpose of the Study:

• To develop and implement a non-parametric method for estimating phylogenetic tree distributions.
• To identify gene trees that significantly deviate from the majority in a given sample.
• To provide a computationally efficient tool for detecting unusual evolutionary events.

Main Methods:

• Proposed and implemented kdetrees, a non-parametric method for estimating tree distributions.
• The method identifies outlying trees by comparing their distributions.
• Implemented as an R package available on CRAN.

Main Results:

• kdetrees shows favorable comparison with existing methods, with improved computational complexity (quadratic in the number of trees).
• Identified unreliable sequence alignments in Apicomplexa genes that were previously undetected.
• Detected a gene with a possible horizontal gene transfer event in Apicomplexa and a paralogy instance in Epichloë.

Conclusions:

• kdetrees is an effective and computationally efficient tool for identifying outlying phylogenetic trees.
• The method aids in discovering biologically significant evolutionary events, including horizontal gene transfer and paralogy.
• kdetrees can help identify problematic sequence alignments requiring further investigation.