SNP genotyping and imputation

This study used genetic markers obtained as described by OConnor, Kilian [44] and briefly outlined here. DNA was extracted from leaves of 295 seedlings and their parents, and sequenced by Diversity Arrays Technology (DArT) Pty Ltd. DArT performed digestion/ligation reactions using a combination of PstI and HhaI restriction enzymes and barcoded adaptors. After PCR, samples were pooled, applied to c-Bot (Illumina) bridge PCR, and then sequenced using Illumina Hiseq2500 for 77cycles. Sequences were processed using proprietary DArT pipelines, with SNP markers detected based on parsing sequence clusters. Missing calls were imputed using the probabilistic principal components analysis (PPCA) method [66] with 97.2% accuracy, which was determined by excluding an additional 10% of missing values and calculating the correlation between the imputed calls and the original dataset. Quality control was performed using pre-imputation parameters, including 50% original call rate, 2.5% minor allele frequency, and a test of Mendelian inconsistencies (parent-offspring trio opposing homozygotes) determined using 16 (50%) of the families. This quality control resulted in 4113 SNPs for genomic analysis.