The Impact of Missing Data on Species Tree Estimation.

Area of Science:

• Evolutionary Biology
• Phylogenetics
• Genomics

Background:

• Phylogeneticists increasingly use large genome-scale datasets with hundreds of genes for clade resolution.
• The impact of moderate to high missing data on species tree estimation remains debated.
• Incomplete lineage sorting (ILS) and gene rate heterogeneity are key factors influencing phylogenetic inference.

Purpose of the Study:

• To assess the effects of missing data on species tree estimation methods.
• To evaluate method performance under varying degrees of incomplete lineage sorting (ILS) and gene rate heterogeneity.
• To determine the influence of random versus nonrandom missing data distribution.

Main Methods:

• Analysis of simulated and empirical phylogenomic datasets.
• Comparison of concatenation (RAxML), gene-tree coalescent (ASTRAL, MP-EST, STAR), and supertree (MRP) methods.
• Assessment across different levels of missing data, ILS, and gene rate heterogeneity.

Main Results:

• Concatenation, coalescent, and supertree methods perform reliably with randomly distributed missing data if enough genes are sampled.
• High missing data with random distribution requires extensive gene sampling when ILS is high or datasets are indecisive.
• Nonrandom missing data significantly challenges STAR and RAxML, especially with high ILS and gene rate heterogeneity.

Conclusions:

• ASTRAL, MP-EST, and MRP methods demonstrate greater robustness to missing data, including nonrandom distributions and high ILS.
• Randomly distributed missing data requires sufficient gene sampling for reliable species tree inference.
• Understanding missing data patterns is crucial for accurate phylogenomic analyses.