Mutational Calling with SomaticSeq

Question

[morovatunc]

Dear all hi,

I would like to know if you guys could help me about starting calling somatic mutations with somaticseq. Right now I am overwhelmed with the data and the program that's why I cannot produce data.

My biggest problem is I am not sure how can I create a new classifier. Should I use the data that was studied in the paper as a ground truth? or directly start off new with my current data.

As I understand from the manual, I have to call mutations with individual programs first. After that I will combine the outputs of the individual callers by somaticseq to create a classifier. Then call mutations with somaticseq algorithm? Once I created this classifier, I don't have to redo the creating classifier part. I have to only call the mutation with created classifier.

I will be working with the data from ICGC and TGCA thats why I have to create a robust pipeline. Since my data is too big, I am confused about the part where I create my classifier. How can I build a classifier which will give me true results for my data which is consisted of +250 samples including tumor and normal pairs?

Any information will be valuable for me. If there is someone who has used this program before PLEASE HELP ME!

Thank you,
Tunc

Answer 1

[ttom]

I had similar questions and these are the information I received from the developer

"SomaticSeq published classifier we put up on gDrive is trained based on Stage 3 of the DREAM Challenge, with the 5-tool classification for SNV (i.e., Mutect, Varscan, JointSNVMix, SomaticSniper, and VarDict) and 3-tool classification for INDEL (Indelocator, Varscan, and Vardict). If your call set did not use those tools, then the classifier is invalid. For the tools that you did you use, it is going to assume that the tools classified them as false positives. If you want to train your own classifier, just make sure the training set and the testing set used the same parameters for all the tools, etc."

You can always check contact them for more details "Fang, Li Tai [[e-mail]]"