CDCOCA: a statistical method to define complexity dependent co-occurring chromosomal aberrations

Background

Copy number alterations (CNA) play a key role in cancer development and progression. In general, more than one CNA can be detected in any given tumor; therefore co-occurring genetic CNA may point to co-operating cancer related genes. Existing methods for co-occurrence evaluation so far have not considered the overall heterogeneity of CNAs per tumor, resulting in a preferential detection of frequent changes with limited specificity for each association owing to the frequently high genetic instability of the samples.

Results

We hypothesize that in cancer some linkage-independent CNAs might display a statistically non-random co-occurrence, and that these CNAs could be of pathogenetic relevance for the respective cancer. We also hypothesize that two CNAs co-occurring in samples with overall few changes (low complexity samples) represent a stronger association then coming from samples with a high number of changes. To verify our hypothesis, we here present a simulation based algorithm CDCOCA (complexity dependent co-occurring chromosomal aberrations). For an initial modeling approach, CNA data for bladder cancer and mantle cell lymphoma at cytogenetic band resolution was obtained from our Progenetix reference database (http://www.progenetix.net) and the CDCOCA was applied to them. A display of ~50 most frequent co-occurrences obtained after p value cut off along with selected cancer associated genes are shown here (Figure 1).

Figure 1

50 most frequent associations obtained after p value cut off of 0.02. 50 most frequent associations plotted using cytoscape. Green circles represent gains, orange represent losses. Red triangles represent apoptotic signaling genes and blue triangles represent TGF-beta receptor signaling genes located to these associations. Magenta triangles represent overlapping genes between both signaling pathways.

Conclusions

Our CDCOCA algorithm has constitutes a new approach to establish statistically significant co-occurring regional genomic imbalances from for example CGH data sets containing at least hundreds of individual copy number profiles. Along with finding CNAs from low/intermediate complexity samples, our algorithm points towards a generally low statistical specificity for co-occurrence of regional CNAs in a CNA rich samples, with a negative impact on pathway modeling approaches based on genomic copy number screening analyses derived from such data.