Independent filtering very different across different contrasts
1
0
Entering edit mode
@danieljcavanaugh-15999
Last seen 3.2 years ago

Hi, I am analyzing an RNAseq experiment that had two different treatments (treated and control) and two time points per treatment (week 2 and week 3), with 3 biological replicates for each treatment/time point. I am interested in determining which genes are differentially expressed between the two treatments at each time point (in other words, comparing week 2 treated samples with week 2 control samples and week 3 treated samples to week 3 control samples).

I have used Galaxy to run DESeq2 separately for each contrast and have kept independent filtering on for both comparisons. For Week 2 data, no genes were filtered out by independent filtering, while for Week 3, all genes with a mean count <~45 across the 6 Week 3 samples have been filtered out, making it so that there are thousands more comparisons for Week 2. It doesn't look like week 2 has a substantial amount of low count genes among those with low p values, so I'm not sure why more genes aren't being filtered out, though I do understand that the independent filtering is supposed to optimize the number of genes that have an adjusted p-value below a given FDR cutoff, so for some reason further filtering is not increasing this number of genes.

Ideally, I'd like to implement a single filtering criterion across all samples so that I can look at whether similar genes are differentially expressed across the two time points. I know it has been recommended in the past to find the minimum filtering threshold across all contrasts and apply that to everything, but in this case I worry that would mean have no filtering which would negatively affect my results for Week 3.

Please advise on whether it would be acceptable to apply a more stringent filtering criterion by either going with the Week 3 filtering threshold for all samples, or using a more arbitrary threshold like a minimum of 10 reads for at least 3 of the samples.

Thanks,
Dan

deseq2 independent filtering • 294 views
ADD COMMENT
0
Entering edit mode
@mikelove
Last seen 2 hours ago
United States

I'd suggest just using a reasonable threshold like, minimum of 10 reads for at least 3 samples. The IF procedure just looks for an optimal threshold, and so that happened to be ~45 for that one contrast, but it's perfectly reasonable to just use a common threshold for all contrasts. The way to do this would be to subset the DESeqDataSet at the start, then run DESeq() and results(dds, independentFiltering=FALSE).

ADD COMMENT

Login before adding your answer.

Traffic: 310 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6