A roadmap to robust discriminant analysis of principal components

Area of Science:

• Population Genetics
• Genomic Analysis
• Bioinformatics

Background:

• Population structure identification is crucial for conservation, wildlife management, and medical genetics.
• Discriminant Analysis of Principal Components (DAPC) is a popular method for differentiating individuals into genetic clusters.
• Effective parameterization of DAPC models is essential for reliable outcomes.

Purpose of the Study:

• To investigate the impact of parameter selection on Discriminant Analysis of Principal Components (DAPC) models.
• To explore the consequences of model over-fitting in DAPC analyses.
• To provide guidelines for evaluating DAPC model results.

Main Methods:

• Utilized simulated genetic data to test DAPC model performance.
• Examined the effects of varying DAPC parameters.
• Assessed model outcomes under different scenarios of genetic variation and gene flow.

Main Results:

• Demonstrated the critical importance of careful parameter selection in DAPC.
• Highlighted potential issues arising from over-parameterization, especially with high gene flow.
• Identified key factors influencing the accuracy and biological relevance of DAPC results.

Conclusions:

• Proper parameterization is vital for robust DAPC models.
• Over-fitting DAPC can lead to misleading conclusions, particularly in admixed populations.
• The study offers essential guidance for biologically relevant population structure inference using DAPC.