Hi,

I'm using reduced model to get genes that vary over different conditions and using degPaterns() to split those genes into groups according to their expression pattern (code bellow). What want to know is in this case, should we use parameter blind=T or blind=F in rlog() ?

Table <- data.frame(sampleName = sampleNames, fileName = sampleFiles, time = time, celltype = celltype, condition=condition)
Table
dds <- DESeqDataSetFromHTSeqCount(sampleTable = Table, design= ~ condition)
ddsHTSeq <- dds[rowSums(counts(dds)) > 1, ]
rld <- rlog(ddsHTSeq, blind=F)

dds_lrt <- DESeq(ddsHTSeq, test="LRT", reduced = ~ 1)

dds_res <- results(dds_lrt, alpha = value)

ddsdatres <- as.data.frame(dds_res)
ddsdatres <- ddsdatres[!is.na(ddsdatres$padj),]
res.sig <- ddsdatres[ddsdatres$padj < value,]
res.sig <- res.sig[order(res.sig$padj),]
rld_mat <- assay(rld)
cluster_rlog <-subset(rld_mat, row.names(rld_mat)%in%row.names(res.sig))
library('DEGreport')
rownames(Table) <- Table[,1]
meta <- as.data.frame(colData(dds_lrt))
clustersA <- degPatterns(cluster_rlog, metadata = meta, time = "condition", col=NULL)

png(paste0(outdir,cell,'.genegroupsA', pvalue,'.png'), res = 300, height = 20*300, width = 20*300, bg = "white")

print(clustersA$plot)

dev.off()

I tend to use blind=FALSE,  because it avoids overestimating the dispersion.