Hi,
In I my design includes covariates+diagnoses. In DESeq2 is VSD normalized for the covariates included in the design? If not, how do I export the normalized counts corrected for the covariates?
Thanks
Nirmala
Hi,
In I my design includes covariates+diagnoses. In DESeq2 is VSD normalized for the covariates included in the design? If not, how do I export the normalized counts corrected for the covariates?
Thanks
Nirmala
No, VST data is not corrected for covariates.
Your question seems to mix VST and normalized counts, which are not the same. VST data are similar to normalized and log-transformed counts, so they can be thought of being on a log scale.
You can correct VST data for covariates by running a simple linear model (VST data~covariates) and taking the residuals (you will lose the information about the mean though). If your covariates are categorical, you can also use ComBat, which implements an empirical Bayes-moderated adjustment which can reduce individual variability at the expense of leaving a residual dependence on the covariate. ComBat is more advanced in that it also corrects the variances, not just means, but it only works with categorical covariates (e.g., batches). WGCNA implements an empirical Bayes-moderated linear model adjustment (empiricalBayesLM) that works with both categorical and continuous covariates but it only corrects means. The empiricalBayesLM function in WGCNA can also be used to run an unmoderated linear model adjustment which preserves the overall mean.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.