The presence of strong genomic waves creates artificial gains and losses in signal intensities for SNP genotyping arrays, and may lead to spurious CNV calls. To reduce the effect of genomic waves on accurate CNV inference, we developed a regression model to correct and adjust for genomic waves (see Methods section). The essence of our adjustment approach lies in building a regression model for each genotyped sample that correlates the signal intensity at a given marker with GC content within a 1 Mb window centered around the marker, and then calculating the residual for each marker in the array as the adjusted signal intensity. This signal adjustment method does not require a training data set for model construction, and can be applied to many different technical platforms.
