question about the -normal-contamination parameter

[panyuwen]

Thank you very much for your efforts in developing and sharing caveman.

Should I use the default values, i.e., -tum-cn-default 5 -normal-contamination 0.1 OR estimate these parameters using ascatNGS? Which one should be better. Thanks very much!

[AndyMenzies]

Unfortunatly that depends on your data. AscatNGS will only run on Whole Genome data, so if you have Whole Exome or Targeted data then use the cn-default.

If you have genomic data, then using the cn-default will do a resaonable job, but if ascat can generate a good and representative profile over your samples then that will help caveman come up with a better answer.

[panyuwen]

Unfortunatly that depends on your data. AscatNGS will only run on Whole Genome data, so if you have Whole Exome or Targeted data then use the cn-default.

If you have genomic data, then using the cn-default will do a resaonable job, but if ascat can generate a good and representative profile over your samples then that will help caveman come up with a better answer.

Thank you very much! Do you have any suggestions about the "contamination" parameter?

I am working on the WGS data. Based on my current tests, it seems that SNVs identified by (cn-default + contamination-default) is subset of those identified by (cn-default + contamination-ascat), which is also subset of (cn-ascat + contamination-ascat), with only dozens of variant difference.

[AndyMenzies]

The aberrant cell fraction reported by Ascat is the inverse of the normal contamination value needed by caveman (ie 80% aberrant cell fraction -> 20% normal contamination), and that should give a good estimation of the level of contamination in your specific samples.

If you don't have an Ascat result then you need to know a little about the kinds of samples you are processing. 10% (0.1) contamination is a good general default but if you are looking at tissue types that are quite diffuse (eg Breast tumours) then a contamination value of 50% or even 80% might be appropriate. Giving it a value close to the reality of your sample will help the algorithm