Explanation about somatic EVS interpretation

[mortunco]

Hi,

I have question regarding the interpretation of somatic EVS score. Based on Strelka2 publication, it was introduced as a phred-like score. So can I take.

Probability of incorrect variant call = 10 ^ (-SomaticEVS/10)

Currently, (based on my previous issues #119 ), I have huge noise in my one cohort. But other published cohorts seems to have their somaticEVS scores called roughly ~16-17. I cannot use simply 16 arbitrarily. Therefore could you please explain me how should I tackle this problem.

Thank you very much for your help,

T.

[tdelhomme]

Hi,

I am also interested in this particular statistic, in my case when I consider that Probability of incorrect variant call P = 10 ^ (-SomaticEVS/10) it seems that P does not follow the expected distribution. Do you have more informations since your post? Thanks!

[mortunco]

I was able to find good quality calls after somatic EVS > 17 which is not suprising but I dont have any way to justify this arbitrary number. Simply phred score explanation is not very useful. Then, I checked couple previously published WGS data(https://www.ncbi.nlm.nih.gov/pubmed/30033370) and I saw that their somaticEVS scores mostly somaticEVS>17. so I think at that time our WGS data was not in very good quality. But if this is the issue, I couldnt understand why strelka called them in the first place.

Sorry couldnt help you too much.

T.