Variable selection in heterogeneous datasets: A truncated-rank sparse linear mixed model with applications to...

Area of Science:

• Genomics
• Statistical Genetics
• Bioinformatics

Background:

• High-dimensional datasets, particularly in biology and medicine, present challenges for traditional variable selection methods.
• Existing methods like Lasso may yield numerous false discoveries when applied to complex, non-independently and identically distributed (non-i.i.d.) data structures.
• Understanding population structure is crucial for accurate variable selection in genetic studies.

Purpose of the Study:

• To develop a robust framework for sparse variable selection in datasets with unknown multiple subpopulations.
• To adaptively correct for population structure without prior knowledge of the sample origins.
• To improve the accuracy of variable selection in high-dimensional biological and medical data.

Main Methods:

• Proposed a unified framework for sparse variable selection.
• Utilized a low-rank linear mixed model to adaptively correct for population structure.
• Developed a method that automatically selects an appropriate covariance structure complexity.

Main Results:

• Demonstrated the effectiveness of the proposed framework through extensive experiments.
• Showcased superior performance compared to existing variable selection methods.
• Successfully applied the method to diverse genomic datasets from plants, mice, and humans.

Conclusions:

• The proposed framework offers an effective solution for variable selection in structured, high-dimensional datasets.
• The method's ability to adaptively handle unknown population structures enhances discovery in genomics.
• This approach advances the analysis of complex biological data, leading to more reliable insights.