differential expression analysis tools for single end reads

Question

[mike]

Hi,

Which tool would be preferable to detect differentially expressed genes between human control vs treated samples? I have more than 4 biological replicates (for each condition) with single end reads of100 bp.

Thanks

[GouthamAtla]

Have you tried anything from your end to find the answer?

[mike]

Yes I have tried cuffdiff2 but I have seen an article where RPKM/FPKM is not consider appropriate way to normalize RNASeq data. EdgeR is another method to detect differentially expressed genes but it gives false positives. I am not sure which one to rely on

[Nicolas Rosewick]

Give a shot to DESeq2 or edgeR

[mike]

would it be appropriate to get intersection of EdgeR and cuffdiff2 and proceed with that set?

[GouthamAtla]

If you want to consider two methods, go for DESeq(2) and edgeR.

[GouthamAtla]

For sure not tuxedo suite. edgeR and DESeq results should be fine. You can perform GO or GSEA analysis and see which one would be more related to your experiment.

[Sukhi Singh]

Why not Tuxedo?

[GouthamAtla]

Tuxedo suite is good for transcript level analysis but for DE genes, edgeR/DESeq2 performs relatively well. cuffdiff2 is more stringent in DE analysis and its not good relatively in estimating variation using the biological replicates. I have shared few links here (may be they are too old by now) Can someone please explain in simple terms how DESeq2 works? that explains the NB based methods and also few discussions compared cuffdiff2 with NB based tools. It's also reported in few publications.

[Sukhi Singh]

Thanks for the links, I have seen the discussion before as well. I should look deeper, for our needs, Tuxedo perfoms well enough.