I have a math question. I am doing differential expression test. My expression data is already log2'd. I would like to perform t-test between control-case for all genes using t.test function in r. I wonder how to correctly calculate fold change?
for example, if t.test result is saved as res. It also calculates means of two groups (case, control). Should I use res$estimate[1] - res$estimate[2]?
I just need to figure out this math question. Anyone could help me?
No need to do that manually. I suggest you invest quality time in reading about RNA-seq differential testing. Check the manuals of DESeq2, edgeR, limma/voom and see what they do. These tools provide end-to-end solutions for differential testing. Custom statistics are neither robust, nor recommended. I say this not doubting your skills but rather because the last, say, 15 years of biostatistics research have produced many methods far superior to naive t-tests on logcounts. It was (and probably still is) an ongoing research process to produce the most meaningful DE testing machinery, and there is no need for you to start from scratch, neither would it be a smart choice to do so.
Since your data are already normalized you probably want the "limma-trend" pipeline, please google for it and previous threads. This has been discussed a lot of times before. Check biostars and support.bioconductor.org for previous threads and be sure to spend a couple of days to really read existing material via google, there is lots of stuff out there.
Thanks for answering! I actually play around metabolites data and I want to manually do t-test and ORA enrichment as benchmark. If 'limma-trend' pipeline has that, I will definitely check it out.
Thanks for answering! I actually play around metabolites data and I want to manually do t-test and ORA enrichment as benchmark. If 'limma-trend' pipeline has that, I will definitely check it out.