SCT data or scale.data to select top highly expressed genes in scRNAseq to compare with VST-transformed values in bulkRNAseq?

[denvercal1234GitHub]

Hi there,

I had a dataset of bulkRNAseq that was processed via DESeq2. For a sample, I selected the top 100 highly expressed genes as below using the vat-transformed values.

vsd <- DESeq2::vst(dds_Exvivo_DatePlSubs, blind = F)
filtered_vsd <- vsd_Tfh[rowMeans(vsd, na.rm = TRUE) > 0, ]
select_vsd <- order(rowMeans(filtered_vsd), decreasing = TRUE)[1:100]

I now want to check whether there was any genes within these 100 top-highly expressed genes by RNAseq that were not detected by scRNAseq.

I don't think I should use scale.data here, but actually scale.data being z-scored/variance-stabilized data would be more equivalent to vst, because SCT data would only be normalized and log (and not variance-stabilized).

My question is would the SCTransform layer of data be appropriate for this purpose, as below? If so, should I use counts(dds_Exvivo_DatePlSubs, normalized = TRUE) instead to select top highly expressed genes in RNAseq to compare?

avg_expr <- rowMeans(GetAssayData(IndividualDatasetMERGED, layer = "data"), na.rm = TRUE)
avg_expr_filtered <- avg_expr[avg_expr > 0]
select_avg <- names(sort(avg_expr_filtered, decreasing = TRUE)[1:100])

Thank you for your help!

Related: https://support.bioconductor.org/p/9161164/#9161168