Hi Michael,

Now that I could enter my full matrix of samples and I ran LRT for each factor in my design, I would like to see in which conditions are my genes of interest being up/downregulated.

I am doing this manually by plotting specific genes using plotCounts. But sometimes I see weird outliers that were not spotted in the initial PCA for all the samples.

I am using the dds object to plot this:

d <- plotCounts(dds, gene="GeneX", intgroup=c("FactorSH", "FactorSTt"), returnData=TRUE)

ggplot(d, aes(x = FactorSH, y = count, color = FactorSTt)) +  geom_point(position=position_jitter(w = 0.1,h = 0)) +
geom_text_repel(aes(label = rownames(d)), max.overlaps = 5) + theme_bw() + ggtitle("GeneX") + 
theme(plot.title = element_text(hjust = 0.5))

If I get signif. results in the LRT test for this gene, but when I plot it I see this, should I discard that result? Or am I just missing one normalization step to visualize those counts?

Thank you.

Laia

You can remove genes like that by filtering your dds object something like this: This will only keep genes where 5 samples have at least 10 counts.

keep <- rowSums(counts(dds) >= 10) >= 5
dds <- dds[keep,]

That sample might behave better after this kind of filtering.