Initial identification of GSVs was carried out using our cnvHap algorithm, which is applicable to data from a wide range of platforms (including Illumina and Affymetrix genotyping arrays, CGH arrays and next-generation sequencing), and which has greatly improved sensitivity and specificity for detection of short GSVs compared to other commonly-used algorithms [19]. To ensure that our analysis mirrored the procedures that led to the original reported associations, we scored only those GSVs that were of a similar type (deletion/duplication) and that spanned the entirety of the GSV region. In addition, to ensure that only high-confidence calls were included, for the shorter candidate regions (those for which we had probe coverage of 6 or fewer probes – see Table 1) we required that a GSV call included a minimum of 3 consecutive probes in all cases, irrespective of the size of the region being analysed.
