Estimating Genome-wide Phylogenies Using Probabilistic Topic Modeling.

Area of Science:

• Genomics
• Computational Biology
• Evolutionary Biology

Background:

• Genome-wide data offers unprecedented resolution for inferring evolutionary history.
• Computational limitations hinder rapid and accurate phylogenetic inference from large genomic datasets.
• Existing methods often require sequence alignment, which can be a bottleneck.

Purpose of the Study:

• To develop and evaluate an alignment-free computational method for inferring genome-wide phylogenies.
• To address the computational constraints in reconstructing evolutionary relationships.
• To provide a robust and efficient tool for phylogenetic analysis.

Main Methods:

• Developed an alignment-free phylogenetic inference method using probabilistic topic modeling (Latent Dirichlet Allocation).
• Extracted 'topic' frequencies from k-mers derived from multilocus DNA sequences.
• Integrated the method into the Python package Topic Contml, using Contml (PHYLIP) for species tree generation.

Main Results:

• Topic Contml demonstrated efficiency and statistical robustness on simulated and empirical datasets.
• Successfully inferred phylogenies from diverse data, including Australian birds, mammals, and PacBio sequences.
• The method effectively handles gaps and unaligned sequences.

Conclusions:

• Topic Contml offers a computationally efficient and statistically robust solution for genome-wide phylogenetic inference.
• The alignment-free approach overcomes limitations of traditional methods.
• The method provides a valuable tool for evolutionary studies across various taxa.