Entering edit mode
6 weeks ago
feather-W
•
0
Hi,
I want to know besides FPKM, TPM, TMM, RUV, DESeq2 normalized counts, etc., if there are other normalized methods? These normalize methods should preferably take other information into account when normalizing, such as the cell's expansion factor. Please help!
Thanks in advance!
What is a "cell expansion factor"? Generally, a good normalization method adjusts for differences in library size and composition. If any other factors need to be regressed, then you can always use this as part of the model during DE analysis, or directly regress it with tools like removeBatchEffect from limma, ComBat-seq from sva, or try to estimate sources of unwanted variation via something like RUVSeq or svaseq.
Hello ATpoint, thanks for your reply!
Now I have an bulk RNA-seq data from lung cancer and there are two groups: 8 week samples and 0 day samples. When I used DESeq2 to analyze differential express genes, in the result of comparing 8 weeks to 0 days, I found that the related genes of epithelial cells were up-regulated, but the related genes of other cells, such as endothelial cells and fibroblasts, were down-regulated.
However, based on the tissue sections, it is currently established that the endothelial cells in the 8-week samples are amplified with fibroblasts, and are multiplied. Therefore, in theory, in the results of difference analysis, the related genes of endothelial cells, fibroblasts and other cells should not be down-regulated.
The means of "cell expansion factor" is the number of any type cells in 8 week samples compare to 0 day samples.
So I wondered if there were other normalize methods that could address this down-regulation of other cell-associated genes.
Thanks!
So a deconvolution problem, see for example https://support.bioconductor.org/p/9150718/ -- but you would need a good estimate of the proportions. Seems like an experiment where you'd better done single-cell RNA-seq.
Ok, thanks for your help very much! Best wishes for you.