Log likelihood ratio test - turn off or remain on

[Jodi Thomas]

Hello,

I have just discovered corset and it looks fantastic for what I would like to do! I have a de novo transcriptome assembly (no genome available) that I would like to use for both downstream differential expression analysis at the gene level (DGE), as well as looking at differential transcript usage (DTU). I see that the log likelihood ratio test is on by default and with the parameter '-I' this ratio test can be turned off. From reading the wiki and published papers the suggestions are to leave the ratio test on to get gene counts for downsteam DGE, but to turn the ratio test off (include '-I') to then go into lace, supertranscripts etc for downstream DTU. If I am to look at both DGE and DTU does it make sense to do two separate corset runs:

1) corset with ratio test on -> DE analysis of genes

2) corset with ratio test off -> lace, supertranscripts etc -> DTU

Or is it better practice to use the same corset run for both downstream DGE and DTU?

Thanks for your help!

Jodi

[Nadia Davidson]

Hi Jodi,

I would try 2) for both the DGE and DTU analysis. The ratio test only slightly affects DGE results in our tests, but can be very bad for DTU. The convenience of doing both together is probably preferable, especially because having the reads mapped to superTranscripts will allow you to manually inspect results in IGV, which is useful also for DGE. You may get some genes being combined together in the same cluster, but this will happen anyway due to assembly errors and would show up at DTU (when it's really DGE). Good luck!

Cheers,

Nadia.

[Jodi Thomas]

Hi Nadia,

Thanks so much for your fast reply - I will do that.

Thanks,

Jodi :)