STAR-Fusion Terra WDL (with Junction File Input)

A modified version of the STAR-Fusion Terra WDL pipeline that adds support for re-running fusion analysis from an existing Chimeric.out.junction file, bypassing the need to repeat STAR alignment.

Overview

The original STAR-Fusion Terra pipeline requires FASTQ files as input, which means re-running fusion analysis requires re-downloading raw data and repeating the full alignment step. This modified version allows you to skip alignment entirely by providing the Chimeric.out.junction file from a prior run as input, saving significant time and compute cost.

Changes from the Original Pipeline

• Added input_chimeric_junction as an optional input (File?) — accepts plain or gzipped (.gz) junction files
• When a junction file is provided, STAR-Fusion is invoked with -J instead of --left_fq/--right_fq, skipping STAR alignment entirely
• BAM, junction, SJ, and STAR log outputs are now optional (File?) since they are only produced during full alignment runs
• Disk size calculation uses junction file size rather than the FASTQ multiplier when running in junction mode

Input Modes

Mode 1: FASTQ Input (original behavior)

Provide one of the following as in the original pipeline:

• left_fq + right_fq — paired FASTQ files
• fastq_pair_tar_gz — a tarball of paired FASTQs

Mode 2: Junction File Input (new)

Provide:

• input_chimeric_junction — the Chimeric.out.junction or Chimeric.out.junction.gz file from a prior STAR-Fusion run

Note: The Chimeric.out.junction.gz file is produced as an output of this pipeline, so it can be fed directly back in for re-analysis runs.

Inputs

| Parameter | Type | Required | Default | Description | |---|---|---|---|---| | sample_id | String | Yes | | Sample identifier | | genome_plug_n_play_tar_gz | File | Yes | | CTAT genome resource library | | left_fq | File? | No | | Left FASTQ file | | right_fq | File? | No | | Right FASTQ file | | fastq_pair_tar_gz | File? | No | | Tarball of paired FASTQs | | input_chimeric_junction | File? | No | | Chimeric.out.junction(.gz) from prior run | | fusion_inspector | String? | No | | inspect or validate | | examine_coding_effect | Boolean | No | false | Annotate coding effect of fusions | | coord_sort_bam | Boolean | No | false | Coordinate-sort the output BAM | | min_FFPM | Float | No | 0.1 | Minimum fusion fragments per million | | docker | String | No | trinityctat/starfusion:latest | Docker image | | num_cpu | Int | No | 12 | Number of CPUs | | memory | String | No | 50G | Memory allocation | | preemptible | Int | No | 2 | Number of preemptible retries | | extra_disk_space | Float | No | 10 | Extra disk space in GB | | use_ssd | Boolean | No | true | Use SSD storage |

Outputs

| Output | Type | Description | |---|---|---| | fusion_predictions | File | STAR-Fusion predictions (gzipped TSV) | | fusion_predictions_abridged | File | Abridged fusion predictions (gzipped TSV) | | junction | File? | Chimeric.out.junction (gzipped) | | bam | File? | Aligned BAM file | | bai | File? | BAM index (if coord-sorted) | | sj | File? | SJ.out.tab splice junction file (gzipped) | | star_log_final | File? | STAR alignment log | | coding_effect | File? | Coding effect annotations | | extract_fusion_reads | Array[File]? | Fusion evidence reads | | fusion_inspector_validate_fusions_abridged | File? | FusionInspector validate results | | fusion_inspector_validate_web | File? | FusionInspector validate web report | | fusion_inspector_inspect_fusions_abridged | File? | FusionInspector inspect results | | fusion_inspector_inspect_web | File? | FusionInspector inspect web report |

Usage on Terra

1. Upload star_fusion_hg38_wf.wdl and star_fusion_workflow.wdl to your Terra workspace via the Workflows tab
2. For a fresh run, populate left_fq/right_fq (or fastq_pair_tar_gz) in your data table as usual
3. For a re-run from an existing junction file, populate input_chimeric_junction in your data table with the Chimeric.out.junction.gz path from the prior run, and leave the FASTQ columns empty

AI Disclaimer

Portions of this pipeline, including the modified WDL code and documentation, were developed with the assistance of Claude Sonnet 4.6 (Anthropic). All AI-generated code has been reviewed for correctness, but users are encouraged to validate pipeline behavior in their own environment before use in production analyses.

Authors

• Brian Haas (bhaas@broadinstitute.org) — Original STAR-Fusion pipeline
• Amy Olex (alolex@vcu.edu) — Junction file input modifications

Credits

This pipeline is a modification of the original STAR-Fusion Terra WDL developed by the Trinity CTAT project. Please cite the original STAR-Fusion publication if you use this pipeline in your work:

Haas BJ, et al. Accuracy assessment of fusion transcript detection via read-mapping and de novo fusion transcript assembly-based methods. Genome Biology, 2019.