Random error with NetMHCII about nextneopi

[mantczakaus]

Hi,

I did some investigation and I think the problem lies in the concurrent runs.

I have submitted the following command five times almost at once and directing results into different folders, here is just an example:

/opt/conda/bin/python /opt/iedb/mhc_ii/mhc_II_binding.py NetMHCIIpan DRB1*13:01 patient1_tumor.25.fa.split_401-600 25 > results_15/result_1131.txt 2> results_15/error_1131.txt

I think it’s because of the concurrent runs because when I run NetMHCIIpan sequentially one by one it all goes well.

I contacted Morten Nielsen who provides technical support for NetMHCIIpan and I'm waiting for his advice but in the meantime I tried to force the pipeline to run pVACseq sequentially instead of in parallel. I tried to do this by using buffer directive in pVACseq process like so:

process 'pVACseq' {

    tag "${meta.sampleName}"

    label 'pVACtools'

    input:
    tuple(
        val(meta),
        path(vep_phased_vcf_gz),
        path(anno_vcf),
        val(hla_types),
        val(tumor_purity),
        path(iedb_install_ok)
    ) from mkPhasedVCF_out_pVACseq_ch0
        .join(vcf_vep_ex_gz, by: [0])
        .combine(hlas.splitText(), by: 0)
        .combine(purity_estimate_ch1, by: 0)
        .combine(iedb_install_out_ch)
		.buffer(1)

But all it did was it did not trigger any of the pVACseq processes. Could you recommend a workaround? Sth that will make pVACseq run one by one and not in parallel for all the HLA alleles?

[riederd]

Hi,
that's interesting. I never hit that issue. I'll have a look into it as well, but right now I'm a bit busy.
Did you try to reduce the number of cpus for pVACseq, e.g.:
change confg/process.config form

    withName:pVACseq {
        cpus = 10
    }

to

    withName:pVACseq {
        cpus = 2
    }

[mantczakaus]

Hi, thanks for coming back to me. The developer of NetMHCIIpan was unable to reproduce, he recommended for me to contact IEDB support - which is what I'm going to do. In the meantime someone from Nextflow community recommended to set maxForks to 1 which worked! pVACseq is run one by one and the pipelien run successfully! Was your recommendation to setting cpus to 1 with the same aim? https://www.nextflow.io/docs/latest/process.html#maxforks

[riederd]

Hi,
I'm glad that the maxForks worked. It is not exactly the same as setting the cpus, but I'd expect that the effect will be similar.