gagneurlab/OUTRIDER

Quick question about number of samples and computeGeneLengths

Closed this issue · 2 comments

AmrR101 commented

I am working on RNAseq data from a small cohort of rare but heterogeneous disease (12 samples). Although I am able to see clear outliers in edgeR/glimma expression plots, OUTRIDER seems to miss these. I am wondering if that infact is due to small number of samples. How do I set expression filters to may be tease out this differences? Also for computeGeneLengths, what are your recommendations for counts processed using UCSC gtf and annotation.

Thanks!

brechtmann commented

Hi AmrR101:

  • it could indeed be your data set size. I would recommend you to find comparable samples, with similar RNA sequencing but with a different disease or even healthy (gtex).

  • second you could try to set the encoding dimension to a smaller value: We are currently implementing a rule that fits well for data sets with many samples. You could try q = 2.
    Or run findEncodingDim(ods, params = 2:5) and then use the best value.

  • maybe this can rescue it but potentially your dataset is just to small for OUTRIDER to work.

c-mertes commented

As this issue did not get any updates for a while I will close it. @AmrR101, if you have any further questions please reopen this issue or open a new one.