How to use sequenza to detect focal homozygous deletions

[kmavromm]

Hi,

I am preparing to work on a project to call CNV events in a number of WES tumor -normal pairs. The purity of those samples based on pathology estimates and other tools (e.g. SCLUST, Battenberg, FACETS etc) is typically 30% to 50%.

Using orthogonal data, for some of those samples, we know that there are focal (<5Mb) homozygous deletions in certain regions of the genome.

I have tried sequenza with a few of those samples playing with the gamma, kmin (grid search from 20 - 100 and 10 - 60 respectively) and break.method options (full vs het).

However, under no combination was I able to detect the known focal deletions and a reasonable segmentation of the genome. In cases where the algorithm is very sensitive (eg gamma 40 and breaks.method 'full') i get thousands of segments, many of which very short, without Cnt, A, B values (these values are set to NA), spread across the whole genome. When the algorithm runs in a less sensitive setting it does not detect focal events. 

It is theoretically possible to detect these events on these data, and what would the suggested parameters be to enable their detection ?