giannimonaco
commented
5 years ago Hello,
Yes, you should just use TPM values and it should be ok the way you did it.
It is normal to obtain negative proportions in some cases. If they are around zero (let's say between 0 to -5%), I would consider them as noise and I would simply take them as 0. Larger negative values, instead, should be taken as a warning message that the samples in your dataset have a composition that deviates substantially from the samples used to generate the ABIS signature matrix. The reason can be one of the following:
- large technical variability
- very different cell types in your samples,
- changes in gene expression profiles due to the condition you are studying (e.g. disease or treatment).
Hence, large negative values are actually giving you an information and are telling you that the ABIS signature matrix is not perfect for your dataset, and in particular for the cell types which produce the negative values. This is also the reason why I chose to use a deconvolution method that does not use constraints as a strategy to avoid negative values in the results.
Best,
Gianni
eegk
Hi, I wonder if there is any other procedure required after TPM normalization,
I attempted to use your package using the formula below over a data-set and obtained negative proportions in some cases.
https://support.bioconductor.org/p/91218/
tpm3 <- function(counts,len) { x <- counts/len return(t(t(x)*1e6/colSums(x))) }
Best,