Do missing data influence the accuracy of divergence-time estimation with BEAST?

Area of Science:

• Evolutionary Biology
• Computational Biology
• Phylogenetics

Background:

• Time-calibrated phylogenies are crucial for evolutionary biology.
• The impact of missing data on divergence dating, especially using the BEAST uncorrelated lognormal method, remains unclear.

Purpose of the Study:

• To test the robustness of the BEAST uncorrelated lognormal method to missing genetic data.
• To evaluate how missing data affects divergence-time estimates in squamate reptiles.

Main Methods:

• Comparing divergence-time estimates from a nearly complete dataset (20 nuclear genes, 32 species) with subsampled matrices.
• Including matrices with complete and incomplete loci, varying degrees of missing data (up to 75%).
• Assessing the impact of the number of fossil calibration points (2 vs. 9).

Main Results:

• Missing data generally had a minor impact on divergence-time estimates (mean error ~5 Myr).
• Even with 75% missing data across 80% of genes, estimates remained largely accurate.
• Errors increased with higher percentages of missing data (up to ~17 Myr).
• Fewer fossil calibrations (2 instead of 9) caused significantly larger errors (~40 Myr), irrespective of missing data.

Conclusions:

• Missing genetic data has a relatively small effect on divergence dating accuracy using BEAST.
• The number and quality of fossil calibrations are far more critical for accurate phylogenetic dating than missing data.
• Researchers can be more confident in using datasets with missing data for time-calibrated phylogenies with BEAST.