filtering lowly expressed exons

Question

[wiscoyogi]

To reduce false positive testing in differential expression, genes with zero or low counts are pre-filtered. I'm wondering what the equivalent is for this with DEXSeq.

the 'minCount' argument in estimateDispersions does not work with the latest available release (3.18) on bioconductor.

It's also recommended in the guide to do filtering, though the actual methods for doing this are not reported.

I can write my own script to do this and I can go off the threshold of 10 per this post though I'm wondering if there's any sort of built-in approach as was indicated?

thanks!

[ATpoint]

I cannot find any documentation on minCount in the function. It might have been removed long time ago. Anyway, usually edgeR::filterByExpr does a good job at design-aware filtering.

[wiscoyogi]

I think the basis for filtering in edgeR vs. DESeq2 are different ( proportion of samples with counts > some threshold vs. total counts) and I'm not sure this applies

[ATpoint]

it all tries to remove genes with counts so low (with certain respect to group size) that they're impossible to realistically be DE in a reliable fashion