An adaptive threshold determination method of feature screening for genomic selection.

Area of Science:

• Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Complex traits are influenced by a small subset of single nucleotide polymorphisms (SNPs) within a vast genome.
• Accurate and efficient selection of these influential SNPs is a significant challenge in genomic studies.
• Traditional feature screening methods often struggle with unclear thresholds for SNP selection.

Purpose of the Study:

• To develop a novel procedure for selecting a parsimonious set of influential SNPs for complex trait prediction.
• To address the limitations of existing feature screening approaches, particularly the issue of unclear thresholds.
• To enhance both prediction accuracy and computational efficiency in genomic selection.

Main Methods:

• Proposed a Backward Elimination Iterative Distance Correlation (BE-IDC) procedure.
• BE-IDC utilizes an adaptive threshold for SNP selection.
• Evaluated the method through six simulations and application to Arabidopsis thaliana genome-wide data.

Main Results:

• The adaptive threshold estimated by BE-IDC consistently outperformed fixed threshold methods in simulations.
• BE-IDC successfully identified four influential SNPs from over 216,000 SNPs in the Arabidopsis thaliana dataset.
• The identified SNPs included the FRIGIDA gene, consistent with findings from traditional methods.

Conclusions:

• The BE-IDC procedure offers a robust approach for selecting SNPs that are crucial for predicting complex traits.
• BE-IDC balances high prediction accuracy with computational efficiency, meeting key demands in genomic selection.
• This method provides a valuable tool for identifying important genetic markers in large-scale genomic datasets.