TrainCnnFilters

TrainCnnFilters

author
Jonn Smith
description
A workflow for training the 1D and 2D CNN filtration methods in GATK.

Inputs

Required

  • bais (Array[File], required): GCS path to index files for the bam files containing the either the mapped reads from which variants were called, or a bam-out from the variant caller that produced the input VCF files.
  • bams (Array[File], required): GCS path to bam files containing the either the mapped reads from which variants were called, or a bam-out from the variant caller that produced the input VCF files.
  • ref_map_file (File, required): table indicating reference sequence and auxillary file locations
  • truth_beds (Array[File], required): GCS path to bed files with confident regions for the given truth_vcfs
  • truth_vcf_indices (Array[File], required): GCS path to index files for VCF files containing validated variant calls ("truth") for the corresponding called variants in vcfs.
  • truth_vcfs (Array[File], required): GCS path to VCF files containing validated variant calls ("truth") for the corresponding called variants in vcfs.
  • vcf_indices (Array[File], required): GCS path to index files for called variants on which to train / test / validate the CNN models.
  • vcfs (Array[File], required): GCS path to VCF files containing called variants on which to train / test / validate the CNN models.

Optional

  • Create1DReferenceTensors.runtime_attr_override (RuntimeAttr?)
  • Create2DReadTensors.runtime_attr_override (RuntimeAttr?)
  • TrainCnn1D.runtime_attr_override (RuntimeAttr?)
  • TrainCnn2D.runtime_attr_override (RuntimeAttr?)

Defaults

  • prefix (String, default="out")
  • TrainCnn1D.optimizer_beta1 (Float, default=0.9)
  • TrainCnn1D.optimizer_beta2 (Float, default=0.999)
  • TrainCnn1D.optimizer_clipnorm (Float, default=1.0)
  • TrainCnn1D.optimizer_epsilon (Float, default=1e-08)
  • TrainCnn1D.optimizer_learning_rate (Float, default=0.0001)
  • TrainCnn2D.optimizer_beta1 (Float, default=0.9)
  • TrainCnn2D.optimizer_beta2 (Float, default=0.999)
  • TrainCnn2D.optimizer_clipnorm (Float, default=1.0)
  • TrainCnn2D.optimizer_epsilon (Float, default=1e-08)

Outputs

  • cnn_1d_tensors (Array[File])
  • cnn_2d_tensors (Array[File])
  • cnn_1d_model_json (File)
  • cnn_1d_model_hd5 (File)
  • cnn_2d_model_json (File)
  • cnn_2d_model_hd5 (File)

Dot Diagram

TrainCnnFilters