Genome-wide association study (GWAS)

A 700K high-density rice array marker dataset was used to run the GWAS (McCouch et al., 2016). In total, 411,066 SNPs were retained after filtering for missing data (< 20%) and minor allele frequency (< 5%). The population structure of the studied accessions was assessed using principal component analysis (PCA) on the constructed genomic relationship matrix (Zheng et al., 2012) ( Figure S1 ). GWAS was conducted in rrBLUP R package (Endelman, 2011) using the linear mixed model described earlier (Dhatt et al., 2021). SNP markers were declared significant using the P-value threshold of –log10(P) > 6.5, based on method of Li and Ji (2005) using effective number of markers (Li and Ji, 2005; Hussain et al., 2020). Manhattan plot and Q-Q plot were created using R package qqman (Turner, 2018). Phenotypic variance (R² ) explained by each SNP was estimated using the mixed.solve () function from the rrBLUP R package (Endelman, 2011) with SNP having variance equal to Kσ²u, where K is the design matrix of SNP and u is the random effect of the SNP. Additionally, R² explained by the locus having all the significant SNPs was estimated using BGLR R package (Pérez and De Los Campos, 2014). For this, all the SNPs were fitted jointly accounting the LD between the markers via a genomic restricted maximum likelihood method (Dhatt et al., 2021).