Data for: Multi-trait/environment sparse genomic prediction using the SFSI R-package

This data set is from CIMMYT’s Global Wheat Program and includes adjusted phenotypic records of grain yield (ton/ha) from n=3,731 wheat (Triticum aestivum) lines evaluated at four environmental conditions (B2I, B5I, MEL, and LHT), and marker data for 9,045 SNPs. This dataset is a subset, corresponding to the lines that have data in the four environments, from a full dataset containing 29,484 lines evaluated at six environments.

Methods
The original data set includes phenotype data from 29,484 wheat lines that were derived from years 2009 through 2016 evaluated at the CIMMYT’s experimental station in Ciudad Obregon, Mexico. Lines were evaluated under six environmental conditions representing a combination of planting system (bed vs. flat, the later referred to as melgas), number of irrigations (2, 5 irrigations or drip irrigation), and sowing date (optimum, late or early planting). Each year, grain yield trials were established in an α-lattice design with three replicates into incomplete blocks. Moisture-standardized grain yield (ton/ha) was measured at each plot. Least-square means by line and environmental condition were obtained using mixed-effects models with a fixed intercept and the random effects of trial, block (within trial), and replicate (within trial).

Lines were genotyped using GBS (Genotyping-by-sequencing) for 42,706 SNP markers. SNPs with more than 70% of missing values and those with minor allele frequency lower than 5% were removed, leading to 9,045 filtered SNPs. The remaining markers that were missing were imputed with the sample mean of lines at the corresponding loci.