Genomewide Markers for Controlling Background Variation in Association Mapping

Current procedures for association mapping in plants account for population structure (Q) and kinship (K). Here I propose an association mapping procedure that uses genomewide markers (G) to account for quantitative trait loci (QTL) on background chromosomes. My objective was to determine if the G and QG models are superior to the K and QK models. I simulated mapping population sizes of N = 384, 768, and 1536 inbreds that belonged to three known subpopulations. The G and QG models showed the best adherence to the significance level (P) specified by the investigator for declaring QTL. Across different genetic models (15 or 30 QTL), population sizes, and P levels, the Q model suffered from a high number of false positives (N-FP). With the K and QK models, a relaxed P level led to a reasonable number of true QTL detected (N-TQ) with N = 384 or 768 but it led to high N-FP with N = 1536. Compared with the K and QK models, the G and QG models had a better balance between high N-TQ and low N-FP. The results strongly indicated that the G and QG models are superior to the K and QK models.