Multiple correspondence discriminant analysis: an application to detect stratification in copy number variation

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Population stratification can confound genetic association studies.
• Copy number alterations (CNAs) are important genetic variations.
• Accurate inference of population structure from genetic data is crucial.

Purpose of the Study:

• To present multiple correspondence analysis (MCA) for correcting population stratification in CNA data.
• To introduce multiple correspondence discriminant analysis (MCDA) for identifying optimal copy number variants (CNVs) that infer population stratification.
• To propose a novel variable ranking method within MCDA using correlation with class directions.

Main Methods:

• Application of Multiple Correspondence Analysis (MCA) for population stratification correction.
• Development and application of Multiple Correspondence Discriminant Analysis (MCDA).
• Utilizing correlation with class directions for variable (CNV) ranking within MCDA.

Main Results:

• A set of 20 CNVs was identified with 98% predictability for inferring population stratification in HapMap populations.
• Variable selection based on centroid ranking within MCDA outperformed traditional principal axes correlation methods.
• MCA effectively corrected for population stratification in CNA data.

Conclusions:

• MCA and MCDA are effective tools for addressing population stratification in CNA data.
• The proposed MCDA method with centroid ranking provides an optimal set of CNVs for population inference.
• This approach enhances the accuracy of genetic analyses by accounting for population structure.