Hi Huma Shehwana

I never performed such analysis by myself, but based on some experience can comment on this.

We use to get this kind of clustering in the behaviors of gene expressions for the fresh and FFPE samples from RNA Seq data due to formalin fixation post effects (degradation, cross-linking, chemical modifications) and by adopting prolonged and / or inappropriate FFPE archiving as well.

Hence still it is a debatable question that the gene expression information obtained from FFPE RNA Seq is similar to that obtained from fresh samples!!!

This no doubt that in fact a lot depends here how we do our wet lab preps, the way you prepare FFPE RNA, its quality and quantity both, which are then applied to library preparations for sequencing. Also, whether your normal (undiseased) samples belong to fresh or from FFPE samples!! because to find and compare somatic mutations we must need healthy tissues and in FFPE case due to pronounced artifacts presence it is very difficult to achieve this and to find such mutations with high confidence even if it is from healthy individual, but FFPE.

If your FFPE RNA is sufficiently preserved then it can ideally depict the picture of transcriptomic similarities to the RNA extracted from fresh tissues. The choice of cDNA enzyme will play a breakthrough role here and I always prefer SS VILO for FFPE RNA as note that degree of degradation among FFPE samples varies significantly.

https://www.thermofisher.com/order/catalog/product/11756050

In FFPE RNA case the length of RNA molecule is always affected, most degraded FFPE samples should be excluded from the analysis during transcriptomic QC phase, a major step to avoid marked clustering later on, as now we have those samples to analyse, which shared substantial transcriptomic similarities and higher correlations of gene expression with the fresh samples.

In literature we can find that people use Fisher's Exact test applied to each pair of FFPE-FF sample for categorical data. And T-test for the differences in the means between the two conditions. Also, since the FFPE have more percentage of PCR duplicates than FF samples so Sambamba tool is good and fast enough to use here. RSeQC package is best to use for gene body coverage, GC content and pairend inner distance metrics, as these are all those factors markedly influenced the gene expression quality in FFPE RNA case and will impose ambiguities in gene expression profiling upon analysis. So just check them as well.

Regards....