A unified local objective function for optimally selecting SNPs on arrays for agricultural genomics applications

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• Traditional SNP array design relied on ad-hoc methods with variable strategies.
• The Multiple-Objective, Local Optimization (MOLO) algorithm was developed for SNP selection, optimizing SNP information (E score) under constraints like MAF, location uniformity (U score), and obligatory SNPs.
• The MOLO algorithm's empirical E score computation has limitations in balancing SNP uniformity and informativeness and doesn't inherently support uniform SNP distribution.

Purpose of the Study:

• To propose a unified local function as an amendment to the MOLO algorithm for optimal SNP selection.
• To develop a method that allows scalable weighting between SNP uniformity (U score) and informativeness (E score).
• To improve SNP array design for better performance across diverse scenarios.

Main Methods:

• Developed a unified local function to amend the MOLO algorithm.
• Incorporated scalable weights to balance the U score and E score.
• Evaluated the new method's performance using imputation concordance rates.
• Tested the approach on six commercial bovine SNP chips.

Main Results:

• The proposed unified local function allows for flexible weighting between SNP uniformity and informativeness.
• Weighting between the U score and E score resulted in a higher imputation concordance rate compared to using either score alone.
• The effectiveness of the enhanced weighting strategy was confirmed through evaluations on commercial bovine SNP chips.

Conclusions:

• The refined MOLO algorithm with a unified local function offers a more robust approach to SNP array design.
• Balancing SNP informativeness and uniform distribution is crucial for improving imputation accuracy.
• The proposed method demonstrates practical utility and improved performance in real-world applications, such as bovine SNP chip design.