Improved Bayesian inference of hybrids using genome sequences.

Area of Science:

• Population Genetics
• Genomic Inference
• Bayesian Statistics

Background:

• Hybrid inference is crucial for understanding population structure and evolutionary processes.
• Existing methods may not fully account for uncertainties in population haplotype frequencies.
• Modeling linkage and recombination is essential for accurate genomic inference.

Purpose of the Study:

• To develop an improved Bayesian hybrid inference method.
• To account for uncertainty in population haplotype frequencies.
• To model linkage and recombination across the genome for two generations.

Main Methods:

• A novel Bayesian hybrid inference framework was developed.
• The method incorporates population haplotype frequency uncertainty.
• It marginalizes over haplotypes while modeling genomic linkage and recombination.

Main Results:

• The new method yields posterior probabilities similar to Chakraborty and Rannala (2023) with large sample sizes.
• For small sample sizes, probabilities are lower, reflecting increased uncertainty.
• Statistical performance, assessed by Receiver Operation Characteristic (ROC) curves, is equivalent to the prior method.

Conclusions:

• The developed Bayesian method offers a robust approach to inferring hybrids and back-crosses.
• It provides a more comprehensive analysis by including haplotype frequency uncertainty.
• The method is applicable to diverse species, demonstrated by its use on kiwifruit, lizard, and puma datasets.