We used a nonoverlapping window approach to test whether the median signal intensity values within each window correlate with particular genomic features, including GC percentage, segmental duplication, gene content, exon content, simple repeat and conserved genomic region. All these features are annotated in the UCSC Genome Browser annotation databases (12,13). The GC percentage data was retrieved from the gc5Base table, the segmental duplication data was retrieved from the genomicSuperDups table, the gene content annotation was retrieved from the refGene table, the exon annotation was also retrieved from the refGene table, the simple repeat annotation was retrieved from the simpleRepeat table and the most conserved genomic region annotation was retrieved from the phastConsElements28way table. We sectioned the genome into 10 kb, 100 kb or 1 Mb nonoverlapping windows; for each window, we calculated the fraction of bases that fall within the annotated regions for each of the genomic features. Windows with less than three SNPs were excluded from analysis. We then calculated the correlation coefficient between the median LRR values and the fraction of annotated bases within each window across the genome.
