DRAGEN CNV Pipeline
The DRAGEN CNV Pipeline accepts alignments in BAM/CRAM format. The alignments are processed through the CNV pipeline along with reference samples based on the application. Feature signals such as read counts, paired end, split reads, and allele frequency are calculated per sample. Artifacts such as GC content and capture kit biases are corrected for prior to normalization. Normalization of the case sample via principal component analysis (PCA) and singular value decomposition (SVD) removes noise from the sample’s signal. Segmentation is performed on the normalized sample prior to CNV calling and statistical genotyping. Various stages of filtering can be applied to mitigate false positives before emitting the final calls in a standard VCF file as output.