Assume I want to run a simple machine learning of healthy/non healthy classification, how can I use TCGA where it has only data of people with cancer. What do usually people do for example for study of gene expressions?

  • $\begingroup$ Depends. What do you want to study? Genome architecture? Gene expression? Cell structure? Metabolic rate? Variants? Usually, you would compare the healthy to non healthy sample from the same individual but we can't really answer unless you specify what sort of analysis you want to do. $\endgroup$
    – terdon
    Dec 3 '17 at 17:17
  • $\begingroup$ @terdon gene expression. But TCGA doesn't contain healthy cells of the same person. $\endgroup$
    – 0x90
    Dec 3 '17 at 17:18
  • $\begingroup$ @terdon my final goal is to do metabolic analysis, but I need to do gene expression first. $\endgroup$
    – 0x90
    Dec 3 '17 at 17:20
  • $\begingroup$ In cancer research with gene expression profiling, you can find molecular subtypes with machine learning. These subtypes are usually not seen with the standard pathological methods. Here is a famous study where they define a different subtype of breast cancer, that respond differently to therapy. With this gene signature they can identify patients that do not benefit from a certain therapy, but would survive without it (you save them a lot of side effects which come with the therapy). $\endgroup$
    – benn
    Dec 3 '17 at 21:11
  • 1
    $\begingroup$ This was an analysis for my master. See Figure 3.A. I didn't download the data, which was provided as a summarized experiment object of Bioconductor. But I would expect to be downloadable somehow. $\endgroup$
    – llrs
    Dec 4 '17 at 9:01

In the TCGA datasets there is a variable (type) containing the information if the sample was from the tumor or from an adjacent region, which is usually considered as healthy.

As this samples come from the same patient and the same tissues they are used in the analysis for the comparisons and classifications.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.