My task is to create a classification model to determine a disease group from a healthy control using gene expression values. The only data that I've been given is expression values from mmseq for the disease group.
My thought was, since I only have the disease group, to use publicly available RNA-seq or microarray data from GEO that has both disease and control groups to help build the classifier.
Because I don't have access to the raw data or count values for the original RNA-seq experiment, it seems I can't use DESeq2 or edgeR to select differentially expressed genes for the classifier, and since I don't have access to mmseq's full output, I cannot run mmdiff, either. Also, from what I've been reading, it's inappropriate to compare expression values among experiments instead of raw counts.
I'm very confused as to how to proceed; should I try to perform some kind of meta-analysis? I've found this paper that attempts to merge RNA-seq data with microarray data, but even so, it relies on raw count values for the different RNA-seq experiments.
Any help would be appreciated.
Am I missing something here? Your task is to differentiate disease from control, without a control?
Yes, unfortunately. Our collaborator who performs the RNA-seq was supposed to provide us with both disease and control data, but they have only given us data from the disease group. Because of this, my PI has instructed me to just use publicly available data as a control.