Detecting Cancer Somatic Mutations From Multi-Sample Low-Pass Wgs Data

Question

[User 2032]

Hi All,

Q1. I'm trying to find recurrent mutations (SNPs and indels) from multi-sample (~100) low-pass (3~4x) whole genome sequencing data for a certain cancer type. They are all paired with matched normal samples. What's the best strategy to call somatic mutations from this type of data?

Q2. When calling somatic mutations, do you normally filter out calls from repeat regions?

Thanks in advance.

Answer 1

[Sean Davis]

Normally, for somatic mutations, you are going to want enough reads in the tumor and normal to be certain that:

• You can see the variant in the tumor, potentially in the presence of normal tissue admixture.
• You need to have enough reads in your normal to be fairly sure that any variants found in the tumor are not in the normal.
  

Both of these require that you have enough reads to make calls without borrowing information between samples. Therefore, I am not sure that low-pass sequencing data is the best way to go here. I'd be very interested to hear what others think on the issue, though.