What samples can be used for a Mutect2 Panel of Normals

Question

I'm working on calling somatic variants from solid tissue tumors. I have one normal and one tumor sample for each patient. I plan to use Mutect2 to call somatic variants after preprocessing the data with the GATK Best Practices pipeline. Mutect2 recommends the use of a Panel of Normals (PON), but I can't figure out what samples should be used to built the PON. The guide says

Additionally, the samples should come from subjects that were young and healthy to minimize the chance of using as normal a sample from someone who has an undiagnosed tumor. Normals are typically derived from blood samples.

My normal samples are from solid tissue, not blood. Can I use the normal samples from each patient to build the panel of normals? Or would this cause problems because I'm using reusing the same normals for both the PON and as the normal sample for somatic variant calling? My first thought is that most somatic variants in individual normal samples should be filtered out by pooling all the normals together to build the PON but I wasn't sure if there are other reasons to avoid using non-blood based normal samples to build the PON

Answer 1

As suggested by @burger in their comment, you can indeed use solid tissue samples as normals for your PON. You are clearly aware, but I want to spell this out for others, that there are a few caveats that go with the practice of using solid tissue normals rather than blood normals. Mainly, if your normals are "tissue-adjacent", ie sampled from an area of tissue adjacent to the tumor, there is a higher risk of including contaminating tumor cells compared to blood normals. As a result you might have real somatic variants that end up filtered because they are seen in the normal. Though that typically acts more at the level of the direct Tumor-Normal comparison than at the level of the PON, since the PON draws from many normal samples and the chance of seeing the same contaminating somatic mutation in many such samples should be quite low. So ultimately that should be fine.

It's a very common practice to use whatever matched normals you have available to build your PON, and there doesn't seem to be any drawback to doing that as long as you have enough samples (at least a few dozen, ideally a few hundred or more). If you don't have enough samples in your own study, then you should seek to use other samples from young and healthy people that have been sequenced using the same technology (same library prep kits/protocols, same chemistry version etc) since the goal of the PON is to weed out recurrent technical artifacts.