I want to do some differential expression analysis for cancer data. I have found the cancer data from TCGA which is hosted by GDC. Unfortunately, the data doesn't include normal sample. All data comes from cancer sample. I tried to find another data from the cancer I want to analyze which include normal sample but I couldn't find some. The TCGA data use Illumina Hiseq platofrm.

On the other hand, I found some NGS data hosted by NCBI GEO data set which publish normal sample for the same tissue but different experiment and it is not cancer analysis. This experiment use ABSolid2 platform.

My question are:

1. Does it make sense to "integrate" normal sample and cancer sample which comes from different experiment for differential expression analysis?

2. Can I compare sequencing result from 2 different platform with DESeq2 if it makes sense to "integrate" these 2 data? What I think possible it to use FPKM/RPKM as a way to compare because FPKM/RPKM has been normalized. DESeq2 can not use FPKM/RPKM data, only read count data. That's why I ask this question.

I would say that you definitely cannot make comparisons between two very different technologies like this. Illumina's and ABI's sequencing technologies almost certainly have completely different biases, so it will be impossible to know whether any differences you observe are due to differential expression or biases in the technology. This is often true even when comparing experiments from different labs using the same sequencing technology, and the problem is likely much worse here.