First, we note that all tested algorithms were able to detect large-scale genomic aberrations ranging from a 14 Mb deletion to a whole chromosome triplication. We therefore conclude that the SNP-Array under study can be used in cytogenetic research. Yet, as depicted in , CNV-finding algorithms may vary considerably in sensitivity and specificity 
and therefore we recommend a combination of different algorithms to facilitate interpretation of findings. Given this varying accuracy of different CNV-detection algorithms in large-scale genomic aberrations, cautiousness is indicated in interpreting findings from whole genome CNV screenings based on SNP-genotyping data. Common CNVs can comprise only few markers decreasing significantly the signal-to-noise ratio as compared to large-scale genomic aberrations. In order to improve reliability of CNV detection in single individuals, we suggest that detection algorithms should be capable of exploiting prior knowledge about CNV base rates in a given genomic region. To provide information on localization of CNVRs and probability measures of CNV occurrence contained therein, we set out to create standard maps of common CNVRs. Given that CNV-formation rate is considerably higher than the mutation rate for single base pairs 
, it is advisable to create population-specific priors to account for varying degrees of genetic diversity.
To decrease the probability that false positive CNV-calls are used for generating the CNVR-maps, we only took concordant CNV-calls of two independent algorithms into account that overlapped in at least two individuals. On the other hand we screened hundreds of individuals, thus increasing sensitivity.
The maps created this way contain information on CNV hotspots and frequency distributions of CNVs. They represent first drafts of population-specific maps as well as a cross-populational map, probably comprising phylogenetically older CNVRs. We report a substantial overlap with CNVR-regions created from previously published data and also were able to validate our maps using different array types, which vary in resolution and their mode of operation due to different chip-architectures.
We argue, that detection accuracy of CNVs can benefit from information on base rates of CNVs in the general population as provided in our CNVR-maps. Two ways to integrate prior knowledge about common CNVRs are plausible: On the one hand, the restriction of CNV-analysis to markers known to be located in CNVRs allows the use of computationally more expensive algorithms that could outperform current methods. On the other hand information gained from CNVR-maps can serve as prior probabilities in Bayesian terms that can be incorporated in CNV-detection algorithms. In order to alleviate assessing how common CNVs relate to variation in phenotypes, higher sensitivity and specificity rates in CNV-detection are needed. We suggest that the use of prior knowledge about CNVRs as provided by our maps, could yield the necessary refinement of methods. We note that our CNVR-maps do not depict structural variations on the p-arms of chromosome 13, 14, 15, 21 and 22 due to lack of markers in these areas on the Affymetrix 6.0 SNP Array. These regions only contain non-chromosome specific highly repetitive DNA sequences and are usually not covered by microarrays. It is important to stress that the array design for the Affymetrix Human SNP Array 6.0 relied mainly on variation detected in non-African populations an thus our Rw-map might underestimate the total variation present in the Rwandese sample. Specially designed custom arrays that take population specificity into account by making use of information provided by large scale sequencing endeavors such as the 1000 Genomes Project 
will be capable of yielding more accurate estimations on the amount of common copy number variation. This is especially important for Sub-Saharan African populations that show higher degrees of genetic variation 
. In order to render the calling of single CNVs more accurate and to overcome the limitations of the current genotyping platforms, it is mandatory to identify population-specific CNVRs and to enrich the marker density for these known CNVRs on the respective custom arrays.
In summary, we generated comprehensive CNVR-maps using micro-arrays in two cohorts of ethnically distinct individuals from Switzerland and Sub-Saharan Africa. The maps represent an extendable framework that can leverage the detection of common CNVs and additionally assist in interpreting CNV-based association studies.