Variable Selection Separates GWAS Signals of Complex Traits in Citrus ( <i>Citrus</i> spp.)

ABSTRACT Genome‐wide association studies (GWAS) have identified numerous genomic regions associated with agronomically important traits in perennial fruit crops, including citrus ( Citrus spp.). Although GWAS are highly successful, they have limitations in separating and prioritizing significant single nucleotide polymorphisms (SNPs) due to the complex patterns of linkage disequilibrium (LD) within associated regions. To handle complex LD patterns and determine the number of causal SNPs, we performed GWAS followed by ‘sum of single effects’ (SuSiE) linear regression, a variable selection method that separates overlapping significant SNPs. We evaluated 9 fruit‐quality traits of 466 citrus accessions over 4 years and obtained 6887 SNPs by genotyping‐by‐sequencing. Using these phenotype and genotype records, GWAS identified 15 genomic regions associated with these traits. In these regions, SuSiE linear regression successfully separated overlapping significant SNPs and identified seven additional GWAS signals in five regions, providing evidence for multiple independent causal SNPs. These results demonstrate that the variable selection method implemented in SuSiE linear regression enables a more precise dissection of genetic architectures and helps identify independent causal SNP underlying complex traits in citrus.