I am running GATK for the first time and working with RNA-seq data from the MDA-MB-231 cell line grown under two different conditions, with three biological replicates per condition. My goal is to identify condition-specific (differential) variants.

I processed all samples individually following GATK best practices for RNA-seq, including SplitNCigarReads and base quality score recalibration. I then performed somatic variant calling using Mutect2 in tumor–normal (paired, multi-sample) mode, using the publicly available panel of normals (PoN) and germline resource provided here: https://console.cloud.google.com/storage/browser/gatk-best-practices/somatic-hg38

I noticed that GATK also recommends the option of creating a custom panel of normals. In this context, would it be preferable to generate my own PoN (for example, using replicates of one condition), or is the publicly available PoN sufficient for RNA-seq–based variant calling in a cell line?

Additionally, are there any extra considerations or recommended downstream steps I should be aware of when using paired mode for identifying condition-specific variants from RNA-seq data?