Mutect2 (v4.3.0.0)

Description: Calls somatic short mutations — single nucleotide variants (SNVs) and small insertions/deletions (indels) — from tumor sequencing data, with or without a matched normal sample, via local assembly of haplotypes using the GATK Mutect2 Bayesian somatic genotyping engine. Authors: Broad Institute; GenePattern Team, UC San Diego Contact: https://groups.google.com/forum/#!forum/genepattern-help Algorithm Version: GATK 4.3.0.0

---

Summary

Mutect2 is a GenePattern module wrapping the GATK (Genome Analysis Toolkit) Mutect2 somatic variant caller. It detects somatic mutations — genetic changes present in tumor tissue but absent from the germline — by performing local de novo haplotype assembly over active genomic regions and scoring candidate variants with a Bayesian somatic genotyping model.

What problem does it solve?

Identifying somatic mutations (mutations acquired during tumor development, as opposed to inherited germline variants) is a central task in cancer genomics. Standard germline variant callers are not appropriate for this task because somatic mutations are often present at low allele fractions, may be heterogeneous across a tumor, and arise against a background of technical noise and germline variation. Mutect2 is specifically designed to overcome these challenges.

How does it work?

1. Active region detection: Mutect2 scans the input BAM file(s) for genomic regions showing evidence of variation. Only these "active regions" are processed in depth.
2. Local assembly: Within each active region, the tool assembles the reads into a set of candidate haplotypes using a De Bruijn graph.
3. Likelihood scoring: Each read is aligned to each candidate haplotype using a pair hidden Markov model (PairHMM), and the genotype likelihoods are computed.
4. Somatic genotyping: A Bayesian model evaluates the evidence for each variant being somatic rather than germline, using optional population allele frequency priors (from a germline resource) and matched normal data.
5. Filtering (optional): After calling, FilterMutectCalls can be run to annotate variants as PASS or assign one or more failure reasons, leveraging read orientation artifact models (e.g., for FFPE samples), contamination estimates, and statistical filters.

When to use this module

• Whole-genome sequencing (WGS) or whole-exome sequencing (WES) of tumor-normal pairs
• Tumor-only somatic calling (no matched normal available)
• Mitochondrial variant calling / heteroplasmy detection
• Construction of a Panel of Normals (PoN) — run Mutect2 in single-sample mode on a set of normal samples, then use GenomicsDBImport and CreateSomaticPanelOfNormals

Companion index file handling

The module wrapper automatically checks for required companion index files (.bai for BAM files; .tbi for VCF.gz files; .fai and .dict for the reference FASTA). If any index is missing, the wrapper generates it on-the-fly using the appropriate GATK/samtools utility (gatk BuildBamIndex, gatk IndexFeatureFile, samtools faidx, gatk CreateSequenceDictionary) before running the main analysis.

---

References

1. Benjamin, D. et al. (2019). Calling Somatic SNVs and Indels with Mutect2. bioRxiv. https://doi.org/10.1101/861054
2. McKenna, A. et al. (2010). The Genome Analysis Toolkit: A MapReduce framework for analyzing next-generation DNA sequencing data. Genome Research, 20(9):1297–1303. https://doi.org/10.1101/gr.107524.110
3. Van der Auwera, G.A. & O'Connor, B.D. (2020). Genomics in the Cloud: Using Docker, GATK, and WDL in Terra. O'Reilly Media.
4. GATK Mutect2 Documentation: https://gatk.broadinstitute.org/hc/en-us/articles/360037593851-Mutect2
5. GATK Best Practices for Somatic SNVs/Indels: https://gatk.broadinstitute.org/hc/en-us/articles/360035894731

---

Source Links

• GATK Source Repository (GitHub)
• GenePattern Mutect2 Docker Image
• GenePattern Mutect2 Dockerfile

---

Parameters

Name	Description	Default Value
tumor.bam *	Tumor sample BAM (or CRAM) file with index. The wrapper auto-generates a .bai index if absent.	—
reference.fasta *	Reference genome FASTA file. Companion .fai and .dict files are auto-generated if absent.	—
normal.bam	Matched normal sample BAM (or CRAM) file. If provided, enables tumor-normal calling mode. The wrapper auto-generates a .bai index if absent.	—
output.vcf	Base name for output VCF file(s). Final output will be <output.vcf>_unfiltered.vcf.gz (and _filtered.vcf.gz if filtering is enabled).	output.vcf.gz
germline.resource	Population germline allele frequency VCF.gz (e.g., af-only-gnomad.hg38.vcf.gz). The wrapper auto-generates a .tbi index if absent.	—
panel.of.normals	Panel of Normals (PoN) VCF.gz for artifact filtering (e.g., 1000g_pon.hg38.vcf.gz). The wrapper auto-generates a .tbi index if absent.	—
intervals	Genomic intervals to restrict variant calling (.intervals, .list, .bed, or .interval_list). Strongly recommended for WES/targeted panels.	—
tumor.sample.name	SM tag of the tumor sample as recorded in the BAM @RG header. Auto-detected from BAM header if left blank.	—
normal.sample.name	SM tag of the normal sample as recorded in the BAM @RG header. Auto-detected from BAM header if left blank. Only used in tumor-normal mode.	—
af.of.alleles.not.in.resource	Prior allele fraction assigned to variants not found in the germline resource. Mode-dependent default: 1e-6 (tumor-normal), 5e-8 (tumor-only), 4e-3 (mitochondria).	0.0000025
mitochondria.mode	Enable mitochondrial variant calling mode with specialized LOD thresholds suited for heteroplasmy.	false
max.reads.per.alignment.start	Downsampling cap: maximum reads retained per alignment start position. Increase to ≥200 for amplicon or very high-depth data.	50
callable.depth	Minimum read depth required at a site to be considered callable during filtering.	10
extra.args	Additional GATK Mutect2 command-line arguments passed directly to the tool (e.g., --genotype-germline-sites true).	—

* required

---

Input Files

1. tumor.bam The primary input: a coordinate-sorted BAM (Binary Alignment/Map) or CRAM file containing the sequencing reads from the tumor sample. The file must include at least one read group (@RG) header line with a sample name (SM:) tag. A companion .bai BAM index file is required for random access; if it does not exist alongside the BAM file, the wrapper will automatically generate it by running gatk BuildBamIndex. Accepted formats: .bam, .cram.

2. reference.fasta The reference genome in FASTA format (.fa or .fasta) against which the reads were aligned. Must match the reference used during alignment. Two companion files are required:

   • .fai — FASTA index (generated by samtools faidx if absent)
   • .dict — Sequence dictionary (generated by gatk CreateSequenceDictionary if absent)

   The wrapper will auto-generate either file if it is missing.

3. normal.bam (optional) A coordinate-sorted BAM or CRAM file for the matched normal sample from the same patient. Providing a matched normal dramatically improves somatic specificity by allowing Mutect2 to distinguish somatic mutations from germline variants and sequencing artifacts. As with tumor.bam, a .bai index is auto-generated if absent. Accepted formats: .bam, .cram.

4. germline.resource (optional) A block-compressed, tabix-indexed VCF (.vcf.gz) containing population germline allele frequencies — typically the gnomAD allele-frequency-only VCF (e.g., af-only-gnomad.hg38.vcf.gz from the GATK resource bundle). Must include an AF INFO field. Used to set germline prior probabilities during somatic genotyping. A .tbi index is auto-generated via gatk IndexFeatureFile if absent.

5. panel.of.normals (optional) A block-compressed, tabix-indexed VCF (.vcf.gz) representing a Panel of Normals (PoN): a collection of variant sites observed in multiple normal samples, representing recurrent sequencing and technical artifacts. Any variant found in the PoN is flagged during filtering. The GATK resource bundle provides 1000g_pon.hg38.vcf.gz as a community PoN. A .tbi index is auto-generated via gatk IndexFeatureFile if absent.

6. intervals (optional) A file specifying genomic regions to restrict variant calling. Strongly recommended for whole-exome sequencing (WES) and targeted panel sequencing to reduce runtime and limit off-target calls. Accepted formats: .intervals, .list (one chr:start-end entry per line), .bed, or .interval_list (Picard-format). For WGS without specific targets, this parameter may be omitted.

---

Output Files

1. <output.vcf>_unfiltered.vcf.gz A block-compressed, tabix-indexed VCF containing all candidate somatic variants called by Mutect2, including both true somatic mutations and likely artifacts. Each record includes extensive annotations: allele depths (AD), allele fractions (AF), genotype likelihoods, and supporting read counts. This file is the direct output of the Mutect2 calling step before any post-call filtering.

2. <output.vcf>_unfiltered.vcf.gz.tbi Tabix index for the unfiltered VCF, enabling fast random access by genomic coordinate.

3. <output.vcf>_filtered.vcf.gz (produced when filtering is enabled) The post-filtered VCF produced by FilterMutectCalls. Each variant is annotated with either PASS (considered a high-confidence somatic call) or one or more filter flags (e.g., germline, panel_of_normals, strand_bias, weak_evidence, orientation_bias). Downstream analysis should typically use this file.

4. <output.vcf>_filtered.vcf.gz.tbi (produced when filtering is enabled) Tabix index for the filtered VCF.

5. <output.vcf>_f1r2.tar.gz (produced when orientation bias filtering is enabled) Compressed archive of F1R2 read orientation counts per site, used as input to LearnReadOrientationModel. Present only when the orientation bias filter option is enabled.

6. <output.vcf>_read-orientation-model.tar.gz (produced when orientation bias filtering is enabled) The fitted read orientation artifact model (output of LearnReadOrientationModel), passed to FilterMutectCalls. Present only when the orientation bias filter option is enabled.

7. mutect2_run.log A plain-text log file capturing the standard output and error streams of all GATK commands executed, including any auto-generated index steps. Useful for diagnosing errors or auditing the exact commands run.

---

Example Data

Input:

• Tumor BAM: HG008-T (NIST/GIAB HG008 tumor sample)
• Normal BAM: HG008-N-D (NIST/GIAB HG008 matched normal, DNA)
• Reference: hg38 (GRCh38) — available from the GATK Resource Bundle
• Germline resource: af-only-gnomad.hg38.vcf.gz — available from the GATK Resource Bundle
• Panel of Normals: 1000g_pon.hg38.vcf.gz — available from the GATK Resource Bundle

Output:

• Example filtered VCF: HG008-T_filtered.vcf.gz — somatic SNVs and indels called against HG008-N-D, filtered with FilterMutectCalls

Note: GIAB HG008 tumor-normal data and truth sets are available from the NIST Genome in a Bottle Consortium.

---

Requirements

• Platform: GenePattern server (version 3.9.11 or later recommended)
• Language/Runtime: Java 8+ (provided within the Docker container)
• Docker Image: genepattern/mutect2:latest (based on broadinstitute/gatk:4.3.0.0)
• Memory: Minimum 8 GB RAM recommended; 16–32 GB for WGS or high-depth samples. Adjust java.heap.size accordingly.
• CPU: Multi-core recommended; the PairHMM step scales with --native-pair-hmm-threads.
• Disk: Sufficient scratch space for intermediate BAM processing and VCF output. WGS runs may require 50–200 GB of free disk.
• Input BAM requirements:
  • Must be coordinate-sorted
  • Must contain @RG read group headers with SM: sample name tags
  • Duplicate marking is recommended (e.g., via MarkDuplicates) prior to running Mutect2
  • Base Quality Score Recalibration (BQSR) is recommended for best results
• Reference requirements:
  • Must match the genome build used for alignment (hg38 recommended)
  • .fai and .dict companion files required (auto-generated if absent)

---

License

This GenePattern module wraps the open-source GATK toolkit.

• GATK License: BSD 3-Clause — see https://github.com/broadinstitute/gatk/blob/master/LICENSE.TXT
• GenePattern Module: MIT License — see https://github.com/genepattern/Mutect2/blob/main/LICENSE

---

Version Comments

Version	Release Date	Description
4.3.0.0	2023-06-01	Initial GenePattern module release wrapping GATK Mutect2 v4.3.0.0; includes auto-indexing of BAM, VCF, and FASTA companion files; tumor-normal and tumor-only modes; optional FilterMutectCalls and LearnReadOrientationModel steps.