Variant Medium Nextflow Pipeline by khersameesh24 · Pull Request #3 · TRON-Bioinformatics/VariantMedium

khersameesh24 · 2025-12-14T14:03:48Z

nextflow modules for variant filtering, variant calling (snv/indel)
singularity support for variantmedium modules
samplesheet (csv/tsv)
src/ moved to bin/ to fix relative imports
some nf-core standards implemented but would need more refinements later
variantmedium.sh script - launcher for the pipeline - run info with bash variantmedium.sh --help - updated in README

Copilot

Pull request overview

This PR introduces a comprehensive Nextflow pipeline for the VariantMedium somatic variant caller, enabling automated execution of SNV and INDEL calling workflows with support for both conda and singularity environments.

Key changes:

Nextflow DSL2 modules and workflows for variant filtering and calling
Bash launcher script (variantmedium.sh) for simplified pipeline execution
Code reorganization from src/ to bin/ to fix relative imports
Samplesheet-based input handling (CSV/TSV formats)

Reviewed changes

Copilot reviewed 33 out of 71 changed files in this pull request and generated 10 comments.

Show a summary per file

File	Description
workflows/variantmedium_call_variants.nf	Implements SNV/INDEL variant calling workflow with DenseNet models
workflows/variantmedium_filter_candidates.nf	Orchestrates ExtraTrees-based candidate filtering for SNV/INDEL
workflows/variantmedium_stage_data.nf	Stages reference data and ML models
workflows/variantmedium_prepare_inputs.nf	Prepares input TSV files from samplesheet
variantmedium.sh	Bash launcher script orchestrating all 8 pipeline steps
main.nf	Entry point coordinating workflow execution based on execution_step parameter
nextflow.config	Configuration with conda/singularity profiles and parameter definitions
conf/modules.config	Per-module configuration including publish directories
conf/base.config	Process resource labels and defaults
subworkflows/data_staging/main.nf	Stages references and models through dedicated modules
subworkflows/parse_samplesheet/main.nf	Validates and parses input samplesheet
subworkflows/parameter_validation/main.nf	Validates required parameters
modules/variantmedium/filter/main.nf	ExtraTrees filtering module
modules/variantmedium/call/main.nf	DenseNet variant calling module
modules/prepare_inputs/main.nf	Generates TSV files for downstream tools
modules/stage_refs/main.nf	Downloads and stages reference data
modules/stage_models/main.nf	Downloads and verifies ML model weights
bin/prepare_input_files.py	Python script to generate pipeline input TSVs
bin/filter_candidates.py	Python script for candidate filtering
bin/run_variant_medium.py	Entry point for variant calling
bin/src/*	Refactored Python source code with fixed relative imports
README.md	Updated with new pipeline launcher documentation

Comments suppressed due to low confidence (1)

README.md:80

There's a spelling error in "envirnments". It should be "environments".

reference genome and S07604624 SureSelect Human All Exon V6+UTR from UCSC if you need them.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

ozlemmuslu · 2025-12-15T08:58:33Z

 VariantMedium is a deep learning-based somatic variant caller for matched tumor-normal short-read sequencing data. It integrates machine learning–based filtering and 3D convolutional neural networks to classify candidate sites as somatic, germline, or non-variant, with high sensitivity and robustness across diverse genomic contexts and sample types.

-## Dependencies
+## Dependencies (handleled in the module environments)


what do you mean by this?

the modules now have a environment.yml file from which the env is built on-the-go we do not need any prior installations

we don't even need Nextflow installation?

ozlemmuslu · 2025-12-15T09:00:12Z

 - conda >= 4.4 (miniconda >=23.11.0 recommended)
 - CUDA 11.4 (optional for GPU support)

-## Installation


The user would still need to clone the repository, or?

no need to clone the repo as well, the pipeline can be run as nextflow run TRON-Bioinformatics/VariantMedium [--options]

ozlemmuslu · 2025-12-15T09:01:32Z

+  --profile                   STRING      Nextflow profile name (conda, singularity) [default: conda]
+                                          [Parts of the pipeline may not support singularity - Prefer using conda]
+OPTIONAL ARGUMENTS:
+  --config                    PATH        Path to custom config file (.conf)


the parameters in the config file are essential for running the pipeline, this is a required argument

ibn-salem · 2025-12-15T15:00:47Z

We need a very imple minimal example usage for basic users that come with their bam files and want VCFs with variant calls. Only the minimum needed steps should be as simple as possible explained.

All further detail can come in later section of the README or in a separate documentation.

ozlemmuslu · 2025-12-15T11:17:08Z

+
+    stub:
+    """
+    touch fake.somatic_snv.VariantMedium.tsv


what are these files for?

ozlemmuslu · 2025-12-15T11:29:08Z

+  reference_dir          = 'ref_data'              // directory name for reference data
+  models_dir             = 'models'                // directory name for trained models
+
+  // call


Actually, some of the parameters (learning rate, drop rate, possibly epoch) below are not updated during inference stage, and are only there for data ops. The others (aug_rate and aug_mixes) might really impact the results and should not be changed (esp. aug_rate, aug_mixes is also mostly there for tracking)

could you add a note, that they are not to be changed

ozlemmuslu · 2025-12-15T11:29:36Z

-
 manifest {
  name = 'TRON-Bioinformatics/variantmedium'
  author = 'Ozlem Muslu, Jonas Ibn-Salem, Shaya Akbarinejad, Luis Kress'


Please add your name as well :)

ozlemmuslu · 2025-12-15T11:32:01Z

  nextflowVersion = '>=24.10.3'
-  version = VERSION
-  doi = DOI
+  version = '1.1.0'


Let's update the version to 1.1.1

ozlemmuslu · 2025-12-15T11:43:05Z

+#---------------------------------------
+
+TSV_FOLDER="${OUTDIR}/tsv_folder"
+REF_DIR="${OUTDIR}/data_staging/ref_data"


variables on lines 171-174 should be pulled from the config file

This behaviour (setting a default and updating with config later) is risky because user might think they don't need to input these files, but have BAMs aligned to other versions of the reference genome

ozlemmuslu · 2025-12-15T11:57:20Z

 VariantMedium is a deep learning-based somatic variant caller for matched tumor-normal short-read sequencing data. It integrates machine learning–based filtering and 3D convolutional neural networks to classify candidate sites as somatic, germline, or non-variant, with high sensitivity and robustness across diverse genomic contexts and sample types.

-## Dependencies
+## Dependencies (handleled in the module environments)


we don't even need Nextflow installation?

ozlemmuslu · 2026-01-08T12:40:26Z

-    touch fake.all_scores_somatic_indel.tsv
-    touch fake.all_scores_germline_indel.tsv
+    touch sample.somatic_snv.VariantMedium.tsv
+    touch sample.germline_snv.VariantMedium.tsv


not sure why we have the files with germline here, VariantMedium is not designed for germline variant calling, and I have it only for experimental reasons

khersameesh24 added 21 commits November 12, 2025 15:30

added nf modules for variantmedium python scripts

bd039f8

added nf modules for data staging

3c88445

added modules config, multiqc

ec66ab3

data staging subworkflow

1ee5ad2

added samplesheet handling

5a2e707

fixed relative paths in output tsv files and script refactoring

940a039

testing changes

1f39db8

parameter validation and modularity

77bd015

added variant calling workflow, changes to outdir

7e19a68

added tabix indexing for strelka

92158f5

pipeline testing changes

cf730fd

importable packaging for variantmedium filter

7557dcc

minor changes and updated README

e1f7d64

removed samplesheet

47ba31b

fixes to pipeline launcher script

4ad9148

fixed relative imports for variantmedium

97c600d

testing changes

b4ac2b5

fixed relative imports, other testing changes

0c29c49

testing changes

5f75a04

testing changes

105c847

testing changes

58d197c

khersameesh24 requested review from Copilot, ibn-salem and ozlemmuslu December 14, 2025 14:04

Copilot started reviewing on behalf of khersameesh24 December 14, 2025 14:05 View session

updated README

65d0658

Copilot AI reviewed Dec 14, 2025

View reviewed changes

khersameesh24 added 2 commits December 14, 2025 15:14

removed absolete files

57fc10b

implemented copilot suggestions

7a3a671

ozlemmuslu reviewed Dec 15, 2025

View reviewed changes

ibn-salem reviewed Dec 15, 2025

View reviewed changes

implemented suggestions

014b5bd

ozlemmuslu approved these changes Jan 8, 2026

View reviewed changes

ozlemmuslu merged commit 05685c9 into TRON-Bioinformatics:main Jan 8, 2026

Conversation

khersameesh24 commented Dec 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

khersameesh24 commented Dec 14, 2025 •

edited

Loading