Assessing effectiveness of many-objective evolutionary algorithms for selection of tag SNPs

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) identify genetic disease causes by analyzing single-nucleotide polymorphisms (SNPs).
• Genotyping all SNPs is computationally intensive; tag SNPs offer a representative subset for efficiency.
• Tag SNPs can introduce missing data or errors, necessitating optimized selection strategies.

Purpose of the Study:

• To address the tag SNP selection challenge using many-objective evolutionary algorithms.
• To optimize the trade-offs between minimizing tag SNP count and maximizing data tolerance.

Main Methods:

• Formulated tag SNP selection as a many-objective optimization problem.
• Applied and investigated Nondominated Sorting based Genetic Algorithm III (NSGA-III) and Multi-Objective Evolutionary Algorithm based on Decomposition (MOEA/D).
• Evaluated greedy versus random initialization methods for algorithm performance.

Main Results:

• MOEA/D generally outperformed other algorithms.
• NSGA-III showed superior performance in maximum tolerance rate compared to NSGA-II.
• SPEA2 achieved the best average Hamming distance.

Conclusions:

• Many-objective evolutionary algorithms significantly outperform existing methods for tag SNP selection.
• Greedy initialization demonstrates clear advantages over random initialization for tag SNP selection using NSGA-III, SPEA2, and MOEA/D.