I have raw read counts (Edgeseq technology) of 56 patients who have been ranked with Mandard score as responders and non-responders;

I have done differential expression by DESeq2 and I obtained a list of DEGs between responders Vs non-responders patients; How I can build a model to find genes with most contribution in the difference of **responders and non-responder patients? I mean How I can train a model by my differentially expressed genes to find genes can contribute to stratification of patients into groups of responders and **non-responders.

For example this is mean of expression of top 20 differentially expressed genes in each group; Their expression seems pretty close to each other so I am not sure how to report a strong signature of genes that are able to predict responders and non-responders

  • 1
    $\begingroup$ Could you expand what have you tried to do? You write about a model to classify two groups, think about what does it mean to have a high fold change when comparing these groups. Look for tools to classify samples $\endgroup$ – llrs Jan 10 '19 at 14:38
  • 1
    $\begingroup$ Yes, they do contribute. The problem is when and where to put the line/threshold, and how specific are these genes $\endgroup$ – llrs Jan 10 '19 at 15:51
  • $\begingroup$ Exactly... Which threshold should be a cut-off for selecting these genes for training a model. Do you have any idea please? $\endgroup$ – Exhausted Jan 10 '19 at 16:17
  • $\begingroup$ The least numbers that provide the targeted replicability or statistical power (if such a thing exists for this kind of "tests") by being specific and sensitivity, true positives, true negatives, false positives and false negatives. See also this related article: Most Random Gene Expression Signatures Are Significantly Associated with Breast Cancer Outcome $\endgroup$ – llrs Jan 10 '19 at 16:28
  • 2
    $\begingroup$ Xref: biostars.org/p/357735 $\endgroup$ – Devon Ryan Jan 10 '19 at 17:25

Your Answer

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

Browse other questions tagged or ask your own question.