Selecting a Window Size for Phylogenomic Analyses of Whole Genome Alignments Using AIC

Area of Science:

• Genomics
• Phylogenetics
• Bioinformatics

Background:

• Gene tree discordance complicates phylogenomic analysis and reconstruction of phylogenetic trees.
• Traditional non-overlapping window approaches often use arbitrary fixed window sizes, risking the inclusion of recombination events.

Purpose of the Study:

• To develop an information-theoretic method for selecting optimal window sizes in phylogenomic analyses.
• To address challenges posed by gene tree discordance and missing data in genomic alignments.

Main Methods:

• Simulated chromosome alignments to evaluate window size accuracy.
• Utilized the Akaike Information Criterion (AIC) to predict window size performance.
• Developed a stepwise non-overlapping window approach to handle missing data.

Main Results:

• AIC effectively predicts window size accuracy for recovering tree topologies.
• Optimal window sizes for Heliconius butterflies were <125bp to 250bp; for great apes, 500bp to 1kb.
• Window size significantly impacts phylogenetic inference, with smaller windows sensitive to gene tree error and larger windows to concatenation effects.

Conclusions:

• The proposed AIC-based method offers a less arbitrary approach to selecting window sizes for non-overlapping window analyses.
• Optimal window size selection is crucial for accurate phylogenetic reconstruction in the presence of gene tree discordance and recombination.