Subsampling reveals that unbalanced sampling affects STRUCTURE results in a multi-species dataset

Area of Science:

• Population genetics
• Evolutionary biology
• Bioinformatics

Background:

• Understanding genetic population structure is crucial for evolutionary and demographic studies.
• The STRUCTURE program is a key tool for inferring genetic clusters.
• Unbalanced sampling designs in STRUCTURE can lead to biased inferences.

Purpose of the Study:

• To test the impact of unbalanced sampling on STRUCTURE inferences.
• To evaluate the effectiveness of subsampling versus changing ancestry models to correct for sampling bias.

Main Methods:

• Analysis of genetic data from 12 high-alpine plant species.
• Comparison of STRUCTURE results using full data sets versus subsampled data.
• Assessment of STRUCTURE results with default versus alternative ancestry models.

Main Results:

• Subsampling Alpine populations in four of 12 species significantly altered clustering compared to full data.
• Alternative ancestry models in STRUCTURE produced results indistinguishable from the default model.
• Unbalanced sampling demonstrably affects genetic cluster inference.

Conclusions:

• Subsampling is a more viable strategy than alternative ancestry models for mitigating unbalanced sampling bias in STRUCTURE.
• Caution is advised when interpreting STRUCTURE results from unbalanced sampling designs.
• Subsampling requires prior knowledge of expected population clusters.