Validation of the CNP genotypes from such an approach is important, but currently hampered by the lack of a gold-standard set of reference genotypes (such as HapMap12 has provided for SNPs). The vast majority of CNPs have not been previously genotyped with accuracy demonstrated in a set of reference samples. We created one such reference dataset (on the basis of consistency across two independent studies of 263 HapMap samples) that has few mendelian inconsistencies, conforms to Hardy-Weinberg equilibrium and shows strong concordance to fosmid end-sequenced samples6. However, this specific dataset is inappropriate for validation of Canary, as it would be statistically overfit and therefore inflate measures of performance.
