bio-variant-calling-deepvariant

Version Compatibility

Reference examples tested with: GATK 4.5+, bcftools 1.19+

Before using code patterns, verify installed versions match. If versions differ:

• CLI: <tool> --version then <tool> --help to confirm flags

If code throws ImportError, AttributeError, or TypeError, introspect the installed

package and adapt the example to match the actual API rather than retrying.

DeepVariant Variant Calling

Installation

Goal: Install DeepVariant via Docker or Singularity container.

Approach: Pull the pre-built container image matching the target platform (CPU or GPU).

Docker (Recommended)

docker pull google/deepvariant:1.6.1

# Or with GPU support
docker pull google/deepvariant:1.6.1-gpu

Singularity

singularity pull docker://google/deepvariant:1.6.1

Basic Usage

Goal: Call germline variants from aligned reads using DeepVariant's deep learning model.

Approach: Run the all-in-one run_deepvariant wrapper specifying model type, reference, reads, and output paths.

"Call variants with DeepVariant" → Convert aligned read pileups into image tensors, classify with a CNN, and output genotyped VCF.

One-Step Run (run_deepvariant)

docker run -v "${PWD}:/input" -v "${PWD}/output:/output" \
    google/deepvariant:1.6.1 \
    /opt/deepvariant/bin/run_deepvariant \
    --model_type=WGS \
    --ref=/input/reference.fa \
    --reads=/input/sample.bam \
    --output_vcf=/output/sample.vcf.gz \
    --output_gvcf=/output/sample.g.vcf.gz \
    --num_shards=16

Model Types

| Model | Data Type | Use Case |

|-------|-----------|----------|

| WGS | Illumina WGS | Whole genome sequencing |

| WES | Illumina WES | Whole exome/targeted |

| PACBIO | PacBio HiFi | Long-read HiFi |

| ONT_R104 | ONT R10.4 | Oxford Nanopore |

| HYBRID_PACBIO_ILLUMINA | Mixed | Hybrid assemblies |

Step-by-Step Workflow

Goal: Run DeepVariant in three explicit stages for more control over intermediate outputs.

Approach: Generate pileup image tensors (make_examples), classify with the CNN (call_variants), then merge and genotype (postprocess_variants).

For more control, run each step separately:

Step 1: Make Examples

docker run -v "${PWD}:/data" google/deepvariant:1.6.1 \
    /opt/deepvariant/bin/make_examples \
    --mode calling \
    --ref /data/reference.fa \
    --reads /data/sample.bam \
    --examples /data/examples.tfrecord.gz \
    --gvcf /data/gvcf.tfrecord.gz

Step 2: Call Variants

docker run -v "${PWD}:/data" google/deepvariant:1.6.1 \
    /opt/deepvariant/bin/call_variants \
    --outfile /data/call_variants.tfrecord.gz \
    --examples /data/examples.tfrecord.gz \
    --checkpoint /opt/models/wgs/model.ckpt

Step 3: Postprocess Variants

docker run -v "${PWD}:/data" google/deepvariant:1.6.1 \
    /opt/deepvariant/bin/postprocess_variants \
    --ref /data/reference.fa \
    --infile /data/call_variants.tfrecord.gz \
    --outfile /data/output.vcf.gz \
    --gvcf_outfile /data/output.g.vcf.gz \
    --nonvariant_site_tfrecord_path /data/gvcf.tfrecord.gz

GPU Acceleration

Goal: Speed up DeepVariant inference using GPU hardware.

Approach: Use the GPU-enabled container image with Docker --gpus flag.

docker run --gpus all -v "${PWD}:/data" \
    google/deepvariant:1.6.1-gpu \
    /opt/deepvariant/bin/run_deepvariant \
    --model_type=WGS \
    --ref=/data/reference.fa \
    --reads=/data/sample.bam \
    --output_vcf=/data/output.vcf.gz \
    --num_shards=16

PacBio HiFi Calling

Goal: Call variants from PacBio HiFi long reads.

Approach: Use the PACBIO model type which is trained on HiFi read characteristics.

docker run -v "${PWD}:/data" google/deepvariant:1.6.1 \
    /opt/deepvariant/bin/run_deepvariant \
    --model_type=PACBIO \
    --ref=/data/reference.fa \
    --reads=/data/hifi_aligned.bam \
    --output_vcf=/data/hifi_variants.vcf.gz \
    --num_shards=16

ONT Calling

Goal: Call variants from Oxford Nanopore long reads.

Approach: Use the ONT_R104 model type trained on Nanopore R10.4 chemistry.

docker run -v "${PWD}:/data" google/deepvariant:1.6.1 \
    /opt/deepvariant/bin/run_deepvariant \
    --model_type=ONT_R104 \
    --ref=/data/reference.fa \
    --reads=/data/ont_aligned.bam \
    --output_vcf=/data/ont_variants.vcf.gz \
    --num_shards=16

Exome/Targeted Sequencing

Goal: Call variants from exome or targeted panel data.

Approach: Use WES model type with a BED file restricting calling to target regions.

docker run -v "${PWD}:/data" google/deepvariant:1.6.1 \
    /opt/deepvariant/bin/run_deepvariant \
    --model_type=WES \
    --ref=/data/reference.fa \
    --reads=/data/exome.bam \
    --regions=/data/targets.bed \
    --output_vcf=/data/exome_variants.vcf.gz \
    --num_shards=8

Joint Calling with GLnexus

Goal: Perform joint genotyping across a cohort from DeepVariant gVCFs.

Approach: Generate per-sample gVCFs, then merge and jointly genotype with GLnexus using a DeepVariant-specific config.

For multi-sample cohorts, use gVCFs with GLnexus:

# Generate gVCFs for each sample
for bam in *.bam; do
    sample=$(basename $bam .bam)
    docker run -v "${PWD}:/data" google/deepvariant:1.6.1 \
        /opt/deepvariant/bin/run_deepvariant \
        --model_type=WGS \
        --ref=/data/reference.fa \
        --reads=/data/$bam \
        --output_vcf=/data/${sample}.vcf.gz \
        --output_gvcf=/data/${sample}.g.vcf.gz \
        --num_shards=16
done

# Joint genotyping with GLnexus
docker run -v "${PWD}:/data" quay.io/mlin/glnexus:v1.4.1 \
    /usr/local/bin/glnexus_cli \
    --config DeepVariantWGS \
    /data/*.g.vcf.gz \
    | bcftools view - -Oz -o cohort.vcf.gz

GLnexus Configurations

| Config | Use Case |

|--------|----------|

| DeepVariantWGS | Illumina WGS |

| DeepVariantWES | Illumina exome |

| DeepVariant_unfiltered | Keep all variants |

Output Quality Metrics

Goal: Assess the quality of DeepVariant calls.

Approach: Generate summary statistics with bcftools stats and check Ti/Tv ratio as a quality indicator.

# Variant statistics
bcftools stats output.vcf.gz > stats.txt

# Filter by quality
bcftools view -i 'QUAL>20 && FMT/GQ>20' output.vcf.gz -Oz -o filtered.vcf.gz

# Ti/Tv ratio (expect ~2.0-2.1 for WGS)
bcftools stats output.vcf.gz | grep TSTV

Benchmarking Against Truth Set

Goal: Evaluate DeepVariant accuracy against a GIAB truth set.

Approach: Run hap.py to compute precision, recall, and F1 for SNPs and indels.

# Using hap.py for GIAB benchmarking
docker run -v "${PWD}:/data" jmcdani20/hap.py:latest \
    /opt/hap.py/bin/hap.py \
    /data/HG002_GRCh38_truth.vcf.gz \
    /data/deepvariant_output.vcf.gz \
    -r /data/reference.fa \
    -o /data/benchmark \
    --threads 16

Complete Workflow Script

Goal: Run DeepVariant end-to-end with indexing and statistics in a single script.

Approach: Wrap run_deepvariant, bcftools index, and bcftools stats in a parameterized shell script.

#!/bin/bash
set -euo pipefail

BAM=$1
REFERENCE=$2
OUTPUT_PREFIX=$3
MODEL_TYPE=${4:-WGS}
THREADS=${5:-16}

echo "=== DeepVariant: ${MODEL_TYPE} mode ==="

docker run -v "${PWD}:/data" google/deepvariant:1.6.1 \
    /opt/deepvariant/bin/run_deepvariant \
    --model_type=${MODEL_TYPE} \
    --ref=/data/${REFERENCE} \
    --reads=/data/${BAM} \
    --output_vcf=/data/${OUTPUT_PREFIX}.vcf.gz \
    --output_gvcf=/data/${OUTPUT_PREFIX}.g.vcf.gz \
    --intermediate_results_dir=/data/${OUTPUT_PREFIX}_tmp \
    --num_shards=${THREADS}

echo "=== Indexing ==="
bcftools index -t ${OUTPUT_PREFIX}.vcf.gz
bcftools index -t ${OUTPUT_PREFIX}.g.vcf.gz

echo "=== Statistics ==="
bcftools stats ${OUTPUT_PREFIX}.vcf.gz > ${OUTPUT_PREFIX}_stats.txt

echo "=== Complete ==="
echo "VCF: ${OUTPUT_PREFIX}.vcf.gz"
echo "gVCF: ${OUTPUT_PREFIX}.g.vcf.gz"

Comparison with Other Callers

| Caller | Speed | Accuracy | Best For |

|--------|-------|----------|----------|

| DeepVariant | Moderate | Highest | Production, benchmarking |

| GATK HaplotypeCaller | Moderate | High | GATK ecosystem |

| bcftools | Fast | Good | Quick analysis |

| Clair3 | Fast | High | Long reads |

Resource Requirements

| Data Type | Memory | CPU Time (30x WGS) |

|-----------|--------|-------------------|

| WGS | 64 GB | ~4-6 hours |

| WES | 32 GB | ~30 min |

| With GPU | 32 GB | ~1-2 hours (WGS) |

Related Skills

• variant-calling/gatk-variant-calling - GATK alternative
• variant-calling/variant-calling - bcftools calling
• long-read-sequencing/clair3-variants - Long-read alternative
• variant-calling/filtering-best-practices - Post-calling filtering