kallisto + sleuth: no results

Question

[stephanie.devos]

Hi, I am performing several differential expression experiments with 1-factor, 2-conditions and 3 replicates per condition. I am currently getting a "....killed" message from R every time I run sleuth version sleuth/0.28.0-intel-2015b-R-3.2.1, right around the step so = sleuth_prep(s2c, ~condition)

     so = sleuth_prep(s2c, ~condition) 
reading in kallisto results ...... 
normalizing est_counts 
51842 targets passed the filter 
Killed

Does anyone know how to fix this? Or how to submit the R commands in a script so that I can submit it to our computer cluster, which has much more resources than the ones for the working window?

When I give the same commands, using the more recent sleuth/0.29.0-intel-2017a-R-3.4.0 instead, I get the following errors:

 > so = sleuth_prep(s2c, ~condition)
    reading in kallisto results
    dropping unused factor levels
    ......
    normalizing est_counts
    51842 targets passed the filter
    normalizing tpm
    merging in metadata
    summarizing bootstraps
    ......Error in H5Dread(h5dataset = h5dataset, h5spaceFile = h5spaceFile, h5spaceMem = h5spaceMem,  :
      HDF5. Dataset. Read failed.
    Error in order(indices) : unimplemented type 'list' in 'orderVector1'

Does anyone know what's going on? The most annoying thing is that, a few weeks back, this was working, which could mean resources are more scarse now, or something got updated maybe...? Any tips are welcome. Please note that I am a total R newb

[WouterDeCoster]

The suspense is killing me, I get the impression that your post is incomplete.

[stephanie.devos]

so what do you think?

[Biojl]

Can you paste the full code?

[stephanie.devos]

so what do you think?

[stephanie.devos]

Sure, here it is:

suppressMessages({
   library("sleuth")
 })
sample_id <- dir(file.path("..", "results_sleuthexample"))
sample_id
kal_dirs <- file.path("..", "results_sleuthexample", sample_id, "kallisto")
kal_dirs
s2c <- read.table(file.path("..", "metadata_sleuthexample", "hiseq_info.txt"), header = TRUE, stringsAsFactors=FALSE)
s2c <- dplyr::select(s2c, sample = run_accession, condition)
s2c
s2c <- dplyr::mutate(s2c, path = kal_dirs)
print(s2c)
so = sleuth_prep(s2c, ~condition)

[philgev2]

Hi, Have you fixed the error "Error in order(indices) : unimplemented type 'list' in 'orderVector1'""?

[stephanie.devos]

Hi, no, any idea what it means (why do I get this error with the newer version of Sleuth and not with the old one??) or how I could fix it? As I said, R skills low at the moment, although I try to catch up. Thank you for any suggestion!

Answer 1

[philgev2]

Hi

I was running the code on Amazon EC2. It seems error is because of some broken parallel processing. When i re-run the code with parameter num_cores=2, code run without error.

Thanks Philge

Answer 2

[gilbo]

If anyone comes across this, when I encountered this error I realized I had forgotten to request enough memory from the cluster. I suspect that is also why when someone drops the number of cores it will fix the issue as well.