I am using vst() from DESeq2 and I need to estimate shrunken log2 fold changes between condition A and B(neither of which have any replicates).
vst_data <- varianceStabilizingTransformation(dds)
I am using vst() since low counts have a higher relative variance than high counts and, therefore, making the variance constant should no longer give disproportionate fold changes. Moreover, to my understanding, for high counts, the variance stabilization transformation approaches a value of log2 of the normalised counts.
Thus, does this assume that the vst() has already transformed my count data into a log2 scale allowing me to compute the Log2FC as vst(B) - vst(A) or is it only an approximate logarithm transformation, still requiring me to compute it as log2( vst(B) / vst(A) )?
