paperKB
coga / coga-kb
Help
Sign in

Chunk #13 — CNV analysis

Source
Genome-wide SNP and CNV analysis identifies common and low-frequency variants associated with severe early-onset obesity.
Embedded
yes

Text

Case and control CNVs were called by plate using Birdsuite-1.53 with CEU parameters. Autosomal calls made by Birdseye and Canary with LOD>=10, number of probes>=5, size>=1kb, probe density>=1 per 10kb were kept. Normalized probe intensities produced by apt-probe set-summarize were converted into log2ratios using the plate median as the reference. For each sample, the median-of-absolute-deviation (MAD) and the sum of autocorrelation (SAC) between pairs of probe 1-5 probes apart were calculated from log2ratios, which reflect the level of noise and long range waviness in the data, respectively. Samples yielding excessive CNV calls were filtered out by fitting the number of kept CNV calls per sample as a linear function of the sample’s MAD in samples with a SAC in the lower 90% and removing outliers of the fitted linear model. Adjacent CNVs in the same sample were merged if (i) they were called with the same genotype, (ii) the distance between them was covered by <100 probes with a density > 0.2 probe per kb, (iii) the ratio of the number of probes between them to the number of probes