Multiple cohorts of tumor normal matched samples

[tiger.c...@gmail.com]

Hi Yuchao,

I am using the CODEX to call both germ-line and somatic CNVs. And the tool is easy to use. I like it. However, I do have several questions which I hope you could provide some hints and solutions.

1. I have multiple cohorts of samples, which were sequenced on the same platform but on different dates and different centers. So there must be some batch effects. The depth of coverage among cohorts were marginally different. Do you think it is better to run CODEX on the cohorts separately or to simply combine the cohorts as one?

2. For each subject, I have both normal and tumor samples. I have tried both the normalize/normalize2 functions and the integer/fractional modes. I have noticed that, regardless of what function and mode I chose, the output will always provide CNV callings for all the samples, I mean every normal sample and every tumor samples. While it still seems OK for me, but I supposed that, the integer will provide callings for normals only, and the fractional mode will provide callings for tumors only. Could you please confirm which is what CODEX is designed to do?

Thanks,

Hu Chen

[tiger.c...@gmail.com]

Hi Yuchao,

To make my second question more clear. What confuses me is that, when calling somatic CNVs using the fractional mode, I get roughly 5~10 fractional calls for each normal (blood) sample, taking Chr 7 for example. In the ideal situation, we should not observe any fractional calls for the normal samples, right? It does not bother me that much, since I will only focus on tumor samples.

And for germ-line callings, I find 10 times more callings from the tumor samples than normal samples. Should I use the results in normal samples, or the union of normal and tumor, or the overlapped results of normal and tumor?

Thanks,

Hu Chen