search

Xinglab / rmats-turbo

Other
232 stars 55 forks source link

Error during rMATS analysis in the nextflow #442

Open Motoufiq opened 1 month ago

Motoufiq commented 1 month ago

Hi,

I am running a nextflow pipeline and rMATS is part of the pipeline. All the steps run well except the rMATS step. How should I overcome this errror? See the attached log below for reference. splicing.log.txt rmats_pairs.txt

EricKutschera commented 1 month ago

It looks like this is the error:

 Error in `$<-.data.frame`(`*tmp*`, "Event_type", value = "CA") :
    replacement has 1 row, data has 0
  Calls: get_postprocessing_table -> $<- -> $<-.data.frame

That error happens after rMATS completes. However it looks like most of the reads were filtered out by rMATS:

USED: 182296943
NOT_PAIRED: 5149055
NOT_NH_1: 256246142
NOT_EXPECTED_CIGAR: 21947137
NOT_EXPECTED_READ_LENGTH: 0
NOT_EXPECTED_STRAND: 0
EXON_NOT_MATCHED_TO_ANNOTATION: 688568009
JUNCTION_NOT_MATCHED_TO_ANNOTATION: 592092171
CLIPPED: 0
total: 1746299457

Only 10% (182296943/1746299457) of reads were USED, while 73% ((688568009 + 592092171) / 1746299457) were filtered for NOT_MATCHED_TO_ANNOTATION

I'm not familiar with the pipeline you are running, but my guess is that either the alignment step in the pipeline used different reference files than what was given to rMATS, or your reads just don't align well

Motoufiq commented 4 weeks ago

@EricKutschera thank you. It was something to do with the setting true or false paired_stats. My samples were unpaired.

Motoufiq commented 4 weeks ago

@EricKutschera one question related the parameter setting in the rMATS analysis. If samples are paired then I should --paired-stats is true, if unpaired, then I should set false. Is my understanding correct?

EricKutschera commented 4 weeks ago

If your samples are paired then you can run with --paired-stats, but it's also fine to run without --paired-stats. The default statistical model can compare any two groups of samples. The paired statistical model can use the pairing information to potentially give better results: https://github.com/Xinglab/rmats-turbo/issues/356#issuecomment-1889275667

Motoufiq commented 4 weeks ago

@EricKutschera thank you. My samples are unpaired so I will set --paired-stats as false