DESeq2 or EdgeR or Limma without replication

Hello, I am trying to do different expression using DESeq2, or EdgeR, or Limma but it always obtain error.
Can I use these tool without replication? Cause patient sample, for one condition I only run 1 time.
Help me please!
Thank you so much!

1 Like

Welcome, @ngutrinh

I’m not quite sure what you mean – if you only have a single dataset, then there would be nothing to compare for expression differences (?). Perhaps I am misunderstanding. That is Ok – the short answer is that replicates are required when using these tools in Galaxy.

We found that many usage issues, false errors, and unexpected results were avoided by requiring replicates by default. Even the original tool authors only recommend using the tools without replicates for very limited “exploratory” purposes only – not analysis.

FAQ: Extended Help for Differential Expression Analysis Tools

The FAQ includes a link to the Bioconductor forum where there is much discussion about their strong recommendations about use of replicates with their tools. Punch line: not using replicates produces invalid results (…if the job completes at all…).

I hope that helps!

1 Like

Thank you so much for your reply Jennaj,
I have 3 different groups but each group only have one RNA-seq data, and I tried the tools and got error. Following your answer, each groups should have replicated. Thank you!

1 Like

I have used DESeq2 without any problem. I have triplicates for each factor. I use feature counts to extract the reads. Then I cut the columns to have geneID and reads. Basically 2 columns. Then you can select triplicates for each factor. It was working… Thanks… Nagul

1 Like

Thank you Nagul,
but each factor I have only one sample, no replication, that why it have error. as I know edgeR can do without replication (use bioconductor) but in Galaxy can not do. If there are any solution, please tell me. Thank you so much

1 Like

Thanks Ngutrinh,

Did you try to change the “Fit Type” to local or parametric. I don’t know the difference behind these two. I used the “Mean”. I can try once with single sample and update you whether it works…

Replicates will always be required when running these tools in Galaxy. If you really need to do exploratory non-replicate analysis, then using the tools line-command is the standard alternative.

Apologies – but this requirement in the Galaxy-wrapped version of the tools was necessary to avoid wide-spread odd job errors (that were not only confusing to troubleshoot but draining shared public computational resources, often for no practical purpose) and even more importantly, to alert scientists about problematic experimental design even if/when putatively “successful” green results were produced.

We tried not requiring replicates originally, but very quickly realized it wasn’t working out well for the vast majority of scientists.

Thank you so much for your explanation!

1 Like