Questions About Training Random Forest Model in PTATO

[GexinLiu]

Hi,
I'm currently working on training the random forest model in PTATO and have been using the provided 'training_vcfs' to run the training program. I have encountered some issues and would appreciate your guidance.

I'm unsure how to properly modify the configuration file and execute the training program, as I couldn't find comprehensive instructions for this part. Below are the command and configuration file I used:

Command:

 nextflow run /home/ug2263/software/PTATO/ptato-train.nf \
    -c /home/ug2263/software/PTATO/configs/run_model_demo_test.config \
    --out_dir /home/ug2263/data/LY_OME/01_processing/PTATO_model/model_demo_test \
    -process.memory '800 GB' \
    -process.cpus 64 \
    -process.maxForks 10 \
    -process.queueSize 100 \
    -resume

Configuration File (run_model_demo_test.config):

includeConfig "${projectDir}/configs/process.config"
includeConfig "${projectDir}/configs/nextflow.config"
includeConfig "${projectDir}/configs/resources.config"

params {

  run {
    snvs =true
    QC = false
    svs = false
    indels = false
    cnvs = false
  }

  // TRAINING
  train {
    version = '2.0.0'
  }
  pta_vcfs_dir = '/home/ug2263/data/LY_OME/Training/training_vcfs/TP'
  nopta_vcfs_dir = '/home/ug2263/data/LY_OME/Training/training_vcfs/FP'
  // END TRAINING

  // TESTING
  input_vcfs_dir = '/home/ug2263/data/LY_OME/Training/training_vcfs/TP'
  bams_dir = ''
  // END TESTING

  out_dir = ''
  bulk_names = [
    ['IBFM26', 'IBFM26_shared_filtered'],
    ['PMCCB15', 'PMCCB15_shared_filtered'],
    ['PMCAHH1-FANCCKO', 'PMCAHH1-FANCCKO_shared_filtered'],
    ['IBFM35', 'IBFM35_shared_filtered'],
    ['PB10268', 'PB10268_shared_filtered']
  ]

  snvs {
    rf_rds = ""
  }

  indels {
    rf_rds = ''
    excludeindellist = "${projectDir}/resources/hg38/indels/excludeindellist/PTA_Indel_ExcludeIndellist_normNoGTrenamed.vcf.gz"
  }
  optional {

    germline_vcfs_dir = ''
    callableloci_dir = ''
    autosomal_callable_dir = ''
    walker_vcfs_dir = ''

    short_variants {
      somatic_vcfs_dir = ''
      phased_vcfs_dir = ''
      ab_tables_dir = ''
      context_beds_dir = ''
      features_beds_dir = ''
    }

    snvs {
      rf_tables_dir = ''
      ptato_vcfs_dir = ''
    }

    indels {
      rf_tables_dir = ''
      ptato_vcfs_dir = ''
    }

    qc {
      wgs_metrics_dir = ''
      alignment_summary_metrics_dir = ''
    }

    svs {
      gridss_driver_vcfs_dir = ''
      gridss_unfiltered_vcfs_dir = ''
      gripss_somatic_filtered_vcfs_dir = ''
      gripss_filtered_files_dir = ''
      integrated_sv_files_dir = ''
    }

    cnvs {
      cobalt_ratio_tsv_dir = ''
      cobalt_filtered_readcounts_dir = ''
      baf_filtered_files_dir = ''
    }
  }


}

However, the execution failed. I encountered the following errors:

Error 1:

" N E X T F L O W   ~  version 25.04.6

Launching `/home/ug2263/software/PTATO/ptato-train.nf` [silly_church] DSL2 - revision: d5f55b4f34

WARN: Include with `params()` is deprecated -- pass params as a workflow or process input instead
Cannot find a component with name 'extractInputVcfGzFromDir' in module: /home/ug2263/software/PTATO/NextflowModules/Utils/getFilesFromDir.nf

Did you mean any of these?
  extractInputVcfFromDir


 -- Check script '/home/ug2263/software/PTATO/ptato-train.nf' at line: 8 or see '.nextflow.log' file for more details"

After revising the function ‘extractInputVcfFromDir’ in ptato.nf to address the above error, I encountered another issue:

Error 2
No .vcf(.gz) files found in [/home/ug2263/data/LY_OME/Training/training_vcfs/FP/*/*.{vcf,vcf.gz}].

Upon further investigation, I believe this error originates from 'short_variants.nf' rather than 'getFilesFromDir.nf'. I'm quite confused about this part and would appreciate any solutions or suggestions you might have.

Thank you in advance for your help!

Best regards,
Gexin