Table 1 Definition of predictors.

Length - length of a segment in bases

Segmental duplication - 1 if the candidate is overlapping known region of segmental duplication, 0 otherwise. All regions listed in [32] that could be successfully translated into hg18 by hgLiftOver utility http://genome.ucsc.edu/cgi-bin/hgLiftOver were used, see Additional file 1, Table S2

Closeness to centromere - 1 if the candidate endpoints are within 2 Mb of the centromere, 0 otherwise

Closeness to telomere - 1 if the candidate endpoints are within 2 Mb of the telomere, 0 otherwise

Sign - 1 if the candidate is a gain, -1 if it is a loss

Height - absolute value of the candidate segment mean

Relative height - absolute value of the candidate segment mean divided by the median absolute deviation of the array residuals

Break - absolute di_erence between means of two segments surrounding the candidate divided by the median absolute deviation of the array residuals

Surrounded by Normals - 1 if both surrounding intervals are normals, 0 if one of them is a gain or a loss

Overlap with other patients - factor with levels: GG if there is one or more other patients in the cohort that have overlapping candidates, all of them are gains; LL if there is one or more other patients in the cohort that have overlapping candidates, all of them are losses; GL if there are at least two patients with overlapping candidates, some of them are gains and some are losses; None if there are no other patients with overlapping candidates

Overlap with other patients - percent - proportion of other patients in the cohort that have overlapping candidate

Matching breakpoint in other patients - percent - proportion of other patients in the cohort that have a candidate with at least one exactly matching breakpoint

Close to other candidates - 1 if there is another candidate CNV within 500 kb on the same chromosome in this patient

Percent of Normal - percent of markers on a chromosome where candidate is located that are not lost or gained

Database score of other candidates - average Database score of other candidates on the same chromosome

Overlap with CNAs - number of other patients that have overlapping non-candidate segment of the same sign as the candidate (gain or loss)