Identifying the number of population clusters with structure: problems and solutions

Area of Science:

• Population genetics
• Computational biology
• Ecological modeling

Background:

• The STRUCTURE program is a widely adopted clustering algorithm for analyzing population genetic structure.
• It estimates individual ancestry proportions to infer population clusters, with post hoc analyses often used to determine the number of populations.
• Over 11,500 citations highlight its significance in the field since its 2000 introduction.

Discussion:

• Uneven sampling across populations or hierarchical levels can lead to inaccurate estimations of population clusters by the STRUCTURE algorithm.
• Post hoc analyses, commonly used to determine the number of clusters, are particularly susceptible to bias from uneven sampling.
• This inaccuracy can result in the identification of an incorrect number of population clusters, misrepresenting true genetic structure.

Key Insights:

• Inferences of population structure using the STRUCTURE algorithm can be compromised by uneven sampling strategies.
• The accuracy of determining the number of population clusters is significantly impacted by non-uniform data collection.
• Puechmaille's work highlights critical limitations of standard STRUCTURE analyses under common sampling scenarios.

Outlook:

• Developing robust analytical methods is crucial for accurate population structure inference in the presence of uneven sampling.
• Strategies such as subsampling can mitigate biases introduced by unequal sample sizes across populations.
• Future research should focus on refining computational tools to ensure reliable genetic structure analyses in diverse ecological contexts.