RNA-seq data for binary classification problem

Question

[botloggy]

I have the TCGA-READ RNA-seq data obtained from GDC Portal.

The samples are of types: "primary tumor" and "solid tissue normal" collected from individuals. The solid tissue normal is a normal tissue sample that is adjacent to the primary tumor. Henceforth, the solid tissue normal may not necessarily be a normal tissue as the sample is still from an individual who has a tumor. So, it would be incorrect to label such samples as normal.

For the binary classification problem, I need tumor samples and normal samples from disease and healthy individuals respectively.

I don't know if I am right but I think the normal samples from any given TCGA data that are of blood-derived/ solid tissue normal tissue sample types may not be having samples collected from normal individuals (disease-free).

Can anyone please suggest on where/how do I get the normal samples?

If there is some website with normal samples, how do I match the genes from the current tumor data?

Any suggestions are highly appreciated. Thanks

[ATpoint]

If you want to put them into the same statistical analysis then you are restricted to those TCGA normals since any independent dataset is a completely different experiment so batch effects would dominate any result you generate rather than a true biological effect.

[botloggy]

Thanks for the suggestion.

Answer 1

[igor]

There was a study where they used GTEx normals since TCGA normals are not entirely normal: Comprehensive analysis of normal adjacent to tumor transcriptomes

[botloggy]

Thank you for this link. It is a great resource. Much appreciate it.