Hello!
I am working on differential gene expression of cell free RNA. In this process I have created the count matrix then calculated the log of count per matrix then calculated the Z_score and calculated the variance using log and created these boxplots. So these 2 boxplots are of 2 different datasets. Is it good to go for further differential gene analysis?
It is impossible to tell anything about these plots as code is missing. As said already, standard tests such as t-tests and Wilcox tests are underpowered at low sample size. Take the Wilcox test for example, say in a 4 vs 4 comparison:
> wilcox.test(1:4, 100:104)
Wilcoxon rank sum exact test
data: 1:4 and 100:104
W = 0, p-value = 0.01587
alternative hypothesis: true location shift is not equal to 0
It is barely below the magical 0.05 threshold but if you do this for 10k genes and hit it with multiple testing correction there is no way you ever get anything significant out of it. I do not have an exact literature reference but if you really want to dive in start with https://pubmed.ncbi.nlm.nih.gov/16646809/ and dig forward.
It makes no sense to do custom approaches like using Z-scores if specialized tools with tens of thousands of citations that well stood the test of time exist (DESeq2, limma, edgeR). Please use them to make your life easy and your stats robust.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
If you have and RNA-Seq count matrix, use dedicated tools for DE analysis e.g. DESeq2, edgeR or limma .There are specifically designed for such analysis, other standard statistics test such as t-test are under-powered.
i have used DESeq2 but same issue. if u have a reference for these please share.