Detecting copy-number aberrations with a low false discovery rate

Cost-effective oligonucleotide arrays are still the predominant technique to measure DNA copy number variations (CNVs). However, CNV detection methods for microarrays suffer from a high false discovery rate (FDR) that, consequently, decreases the discovery power of a study after correction for multiple testing and thereby weakens the chance that a genetic association study succeeds. A remedy for suffering from too high FDRs is to filter out putative false detections. We suggest to use a probabilistic latent variable model (cn.FARMS), which is optimized by a Bayesian maximum a posteriori approach, to identify putative false detections by measurement inconsistencies across samples. We rigorously evaluate the performance on both different data sets and microarray platforms. We find that cn.FARMS clearly outperformed the most prevalent methods (dChip, aroma.affymetrix) with respect to FDR and sensitivity, i.e. has fewer false positives while detecting more true CNVs.