AMOVA-based clustering of population genetic data

Area of Science:

• Population genetics
• Computational biology
• Bioinformatics

Background:

• Understanding population genetic structure is crucial for genetic studies.
• Traditional methods like Analysis of Molecular Variance (AMOVA) require a predefined population hierarchy.
• Clustering analyses are often used to infer population structure, but integrating them with F-statistics presents challenges.

Purpose of the Study:

• To develop a unified analytical framework that combines population clustering and F-statistic calculation.
• To demonstrate the theoretical relationship between AMOVA and K-means clustering for population grouping.
• To evaluate the performance of this integrated method for analyzing population genetic data.

Main Methods:

• Developed an AMOVA-based K-means clustering approach to group populations.
• Utilized simulations to test the method under various mating systems (random mating, nonrandom mating, selfing, clonal reproduction) and migration rates.
• Compared two summary statistics, pseudo-F and Bayesian Information Criterion (BIC), for estimating the optimal number of clusters.

Main Results:

• The AMOVA-based K-means clustering method effectively integrates population structure analysis and F-statistic calculation.
• Simulations showed strong performance across different mating systems, with better results under random mating at high migration rates.
• Pseudo-F generally outperformed BIC in estimating the number of clusters, while BIC was more effective at detecting significant genetic structure.

Conclusions:

• The proposed AMOVA-based K-means clustering offers a powerful and flexible tool for analyzing population genetic data.
• This integrated approach overcomes the limitations of a priori hierarchy requirements in traditional AMOVA.
• The method is valuable for researchers investigating population genetic structure and provides downloadable software for implementation.