Multi-objective tag SNPs selection using evolutionary algorithms.

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Single nucleotide polymorphisms (SNPs) and structural variations exhibit linkage disequilibrium, enabling the use of tag SNPs to represent larger sets of genetic variants.
• Tag SNP selection accuracy and power can be compromised by genotyping failures, errors, and allele-specific tagging biases.
• Diverse genotyping platforms and project requirements necessitate tailored tag SNP sets.

Purpose of the Study:

• To formulate tag SNP selection as a multi-objective optimization problem.
• To develop a flexible and efficient method for selecting tag SNP sets that balance multiple criteria.

Main Methods:

• Formulated tag SNP selection as a four-objective optimization problem: minimizing tag SNP count, maximizing missing data tolerance, and balancing allele detection power.
• Employed evolutionary algorithms with greedy initialization to find non-dominated solutions simultaneously addressing all objectives.
• Developed a flexible approach allowing users to extract customized tag SNP sets for various platforms and scenarios.

Main Results:

• The proposed method offers flexibility, enabling the selection of tag SNP sets (e.g., up to 100 tags with 10% missing data tolerance).
• Compared to conventional methods, this approach explores a larger search space and achieves faster convergence.
• Experimental results highlight trade-offs among objectives, indicating that a modest increase in tag SNPs can significantly improve tolerance and power.

Conclusions:

• The evolutionary algorithm provides an efficient and flexible solution for tag SNP selection.
• The method accommodates diverse genotyping needs and project requirements.
• Optimized tag SNP selection is achievable even with low missing and error rates in modern genotyping platforms.