Comparing inference under the multispecies coalescent with and without recombination

Area of Science:

• Evolutionary biology
• Population genetics
• Bioinformatics

Background:

• Accurate inference of population parameters is crucial for understanding evolutionary histories.
• Recombination is a fundamental evolutionary process but challenging to model precisely.
• Existing coalescent methods may introduce biases due to recombination approximations.

Purpose of the Study:

• To evaluate the performance of three population parameter estimation methods: StarBEAST2, SNAPP, and diCal2.
• To assess the impact of varying recombination rates, mutation rates, and incomplete lineage sorting on these methods.
• To compare the accuracy of these methods under different simulation scenarios.

Main Methods:

• Whole-genome simulations were conducted.
• Three distinct inference methods were tested: StarBEAST2, SNAPP, and diCal2.
• Simulations varied key evolutionary parameters including recombination and mutation rates.

Main Results:

• StarBEAST2 demonstrated robustness to realistic recombination rates when using short or medium loci.
• SNAPP was largely unaffected by recombination events, as anticipated.
• Unexpectedly, diCal2, a method explicitly modeling recombination, yielded poorer performance compared to the other methods.

Conclusions:

• StarBEAST2 and SNAPP are reliable for population parameter estimation even with recombination.
• diCal2's performance issues highlight challenges in accurately modeling recombination for phylogenetic inference.
• Further research is needed to improve methods that explicitly account for recombination.