gms | German Medical Science

Besides single nucleotide polymorphisms (SNPs), copy number variations (CNVs) as another important component of genomic variation have gained much attention with regard to human phenotypic diversity. CNVs, being defined as segments of DNA that are larger than 1 kb in size and that are present at variable copy number in comparison with a reference genome, have been discovered to exist on a large scale [1]. Such duplications or deletions of chromosomal segments are believed to be related to various diseases like mental retardation, neurological disorders, or cancers. Any such CNV association analysis is intimately connected with the detection of CNV loci and with the assessment of individual copy number states. To date, SNP genotyping arrays are widely used for CNV identification, although the SNP microarray approaches tend to lead to many false negatives and positives [2]. One striking difficulty when quantifying the copy number signal of each individual under study, is the lack of a canonical reference genome. Several authors, as for instance Oldrige et al. [3], have compared the performce of taking the median values across differently composed groups of samples, supposed to be of normal 2 copy number state, selected from the genotyped study group. Other authors, as for instance Lin et al. [4], suggest to use median or (trimmed) mean values across publicly available reference samples, such as the HapMap samples. On the basis of family-based, genome-wide SNP microarry data (Affymetrix 6.0), we evaluated how a sophisticated per-marker within-sample estimation of reference intensity signals, prior to the application of selected calling algorithms, can influence the precision of CNV detection and consequently also the validity of subsequent association analyses. CNV calling presicion has been judged in comparison to selected qPCR data upraised in the same sample and with regard to Mendelian inconsistency rates.