Macarons

Macarons selects subsets of single nucleotide polymorphisms (SNPs) to improve prediction of complex quantitative phenotypes in genome-wide association studies (GWAS).

Key Features:

• Feature selection: Selects small, complementary subsets of genomic variants for phenotype prediction in GWAS.
• High-dimensional handling: Operates when the number of screened loci substantially exceeds the number of samples.
• Linkage disequilibrium avoidance: Avoids selecting redundant pairs of SNPs that are likely to be in linkage disequilibrium.
• Parameter control: Employs two interpretable parameters to control the trade-off between computational time and predictive performance.
• Speed: Achieves runtimes at least two orders of magnitude faster than comparative selection methods.
• Predictive performance: Demonstrates similar or superior phenotype prediction compared to state-of-the-art selection methods.
• Scalability: Scales to genome-wide variant sets, processing on the order of 10^7 variants within minutes.

Scientific Applications:

• SNP selection for GWAS: Identifying SNP subsets associated with complex, quantitative phenotypes.
• Improved phenotype prediction: Enhancing prediction of quantitative traits by selecting complementary loci that jointly explain phenotypic variation.
• Large-scale genomic analyses: Applying feature selection to genome-scale datasets containing millions to tens of millions of variants.

Methodology:

Selects a small complementary subset of SNPs while avoiding redundant SNP pairs likely in linkage disequilibrium, with selection behavior governed by two interpretable parameters that trade off computational time and performance.