Hello, bioinformatics community

I am currently working on a pan-cancer differential expression (DE) analysis and am undecided about which dataset processing pipeline to use. Here's what I have done so far:

Using TCGA Data from Biolinks:

1. I downloaded data for multiple cancer types via TCGAbiolinks (R).
2. Integrated all cancer types and removed batch effects using ComBat-seq.
3. Performed DE analysis using DESeq2.

Using TOIL Processed Data from Xena:

1. I downloaded TOIL-processed expected counts data from the Xena hub.
2. Since the TOIL data is log2(x+1)-transformed, I reversed the transformation and then performed DE analysis with DESeq2.

Observations:

Interestingly, the differentially expressed genes (DGE) identified from TOIL data were closer to the DGE I obtained from CCLE data (via DepMap), which made me lean towards using the TOIL pipeline for my study.

However, something still bothers me, and I am not confident that this choice is correct.

Are there significant downsides to using TOIL-processed Xena data instead of TCGA Biolinks data for DE analysis?

I am looking for insights or validations to ensure my approach is methodologically sound.

Thank you for your help!