Subgroup detection in genotype data using invariant coordinate selection.

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• Principle Component Analysis (PCA) is the dominant dimension reduction technique for high-throughput genotype data.
• Alternative methods are rarely utilized despite potential limitations of PCA.

Purpose of the Study:

• Introduce and evaluate Invariant Coordinate Selection (ICS) as a dimension reduction method for genotype data.
• Demonstrate ICS's ability to identify subgroups missed by PCA.

Main Methods:

• Applied Invariant Coordinate Selection (ICS) to simulated and real high-throughput genotype datasets.
• Compared ICS performance against PCA and five other dimension reduction techniques.
• Investigated biological relevance of detected subgroups using genotype data.

Main Results:

• ICS successfully identified subgroups in simulated data where PCA failed.
• Two distinct subgroups were detected in a real chicken genotype dataset using ICS.
• Genotype analysis supported the biological significance of the identified subgroups.
• ICS demonstrated superior performance compared to PCA and other methods in specific scenarios.

Conclusions:

• Invariant Coordinate Selection (ICS) is a valuable tool for high-throughput genotype data analysis, complementing PCA.
• ICS can detect population structure and subgroups missed by traditional PCA.
• Implementation of ICS in statistical environments like R adds minimal computational overhead.