Short sequence motifs thought to form non-B-DNA structures may predispose to chromosomal rearrangements27. We tested the hypothesis that primary DNA sequence can predict CNV formation by screening CNV breakpoints for enrichment of 13 published motifs and genomic annotations (Fig. 3a and Supplementary Methods). Two motifs forming non-B-DNA structures were strongly overrepresented at CNV breakpoints (G-quadruplexes P < 10−3, slipped DNA P < 10−3), as were CpGs and a 13-bp motif predictive of recombination hotspots and genome instability in humans28. In the latter case the association seems to be due solely to VNTR containing the hotspot motif (Fig. 3c). Our results indicate that the previous observations of recombination hotspots flanking a few well-characterized highly polymorphic VNTR29, probably reflects a genome-wide association between hotspots and a large subset of VNTR. The known enrichment of G-quadruplexes and CpGs in gene promoters30 may partly explain the enrichment of CNVs we observed in promoters (Supplementary Figs 1.6 and 1.7).
