Researchers have developed an open-source tool that automates the entire workflow of whole-exome sequencing analysis, from raw sequencing reads to annotated variant calls, and runs the most computationally demanding steps several times faster than a conventional step-by-step approach. The tool, called WEAP (Whole Exome Analysis Pipeline), is described in a study published in BMC Methods and was evaluated on whole-exome sequencing data generated from gastric cancer samples. Its central promise is straightforward: a user supplies the path to a directory of FASTQ files and a handful of essential parameters, and the pipeline handles alignment, variant calling and annotation without further intervention and without requiring specialist bioinformatics expertise.
Whole-exome sequencing is one of the most widely used approaches in clinical genomics because it targets the exonic regions of the genome, which make up only one to two percent of the whole genome but contain the protein-coding sequences where many disease-relevant variants reside. Sequencing these regions is a cost-effective way to identify novel candidate genes associated with disease, detect copy number variations in cancer and other diseases, and support precision treatment and predictive medicine. However, the analysis of the resulting data is a multi-step, time-consuming task that typically demands familiarity with a shifting landscape of bioinformatics algorithms, software tools and databases, and errors at any stage can lead to erroneous variant calls.
WEAP addresses this by wrapping the Genome Analysis Toolkit (GATK) best practices workflow into a single automated pipeline. It uses the bwa-mem2 aligner to map paired-end reads to the hg38 reference genome, samtools to convert SAM files to coordinate-sorted BAM files, Picard version 3.0.0 to mark duplicate reads introduced during library preparation, GATK version 4.5.0.0 for base quality score recalibration, variant calling and filtering, and ANNOVAR for gene-based and filter-based annotation. GNU parallel orchestrates the work, running four jobs at a time so that multiple samples are processed concurrently while each individual tool exploits its own native parallelism.
The germline branch of the workflow follows the standard GATK route: base quality score recalibration, variant calling in gVCF mode, joint genotyping, variant quality score recalibration using known variant resources such as dbSNP 138 and high-confidence panels from the 1000 Genomes Project, and hard filtering with the default criteria of the GATK best practices workflow. The somatic branch implements Mutect2 in two modes: a tumor-only mode and a tumor-with-matched-normal mode, both coupled with a germline resource and a panel of normals (PoN) to filter out false-positive somatic calls, followed by GATK FilterMutectCalls. Annotation draws on RefSeq, dbSNP 150, ClinVar, 1000 Genomes, ExAC, gnomAD exome and, for somatic cancer calls, COSMIC, with pre-classified annotations following ACMG guidelines.
The performance gains reported in the study are substantial. In germline variant calling with four samples, the major steps ran 1.5- to 3.6-fold faster in parallel mode than in serial mode: roughly twofold faster alignment, SAM-to-BAM conversion, sorting and duplicate removal, 1.5-fold faster base quality score recalibration, 3.6-fold faster variant calling and threefold faster annotation. Overall, germline variant calling and annotation for the four samples dropped from 5 hours 58 minutes in serial mode to 3 hours 16 minutes in parallel mode, a 45.25 percent reduction in time. The authors note that combining gVCF files with joint genotyping and the subsequent variant quality score recalibration and filtration steps could not be parallelized, because GATK itself does not perform parallel processing during those stages.
The somatic results were even more striking. Creating a panel of normals from 40 whole-exome datasets of blood origin, sequenced on an Illumina NovaSeq 6000 platform, took 402,900 seconds in serial mode but only 130,711 seconds in parallel mode, roughly a threefold speedup; expressed differently, the time fell from about 111 hours 55 minutes to 36 hours 18 minutes, a 67.56 percent reduction. In tumor-only mode, individual steps accelerated by 1.4- to 7.7-fold, with variant filtration showing the largest gain at 7.7-fold, and the total time fell from 15 hours 28 minutes to 4 hours 55 minutes, a 68.17 percent reduction. In the tumor-with-matched-normal mode, the total time dropped from 21 hours 6 minutes to 6 hours 9 minutes, a 70.83 percent reduction, with alignment 2.6-fold faster, SAM-to-BAM conversion 1.8-fold faster, sorting 2.2-fold faster, duplicate removal 3.4-fold faster, Mutect2 variant calling 3.6-fold faster, filtration 4.1-fold faster and annotation 2.9-fold faster.
The evaluation used in-house whole-exome sequencing data from gastric cancer patients, generated with participant consent and after approval by the institutional ethics committee of Civil Hospital Aizawl in Mizoram, India. Four datasets from blood, tumor tissue and adjacent normal tissue were sequenced at 100-fold depth on the Illumina NovaSeq 6000, and the panel of normals was built from 40 datasets prepared with the same library preparation method on the same platform. The pipeline successfully called and annotated both germline and somatic variants from the trimmed FASTQ files. During alignment, one sample used 16 CPU threads, so four samples in parallel consumed 64 threads; HaplotypeCaller and Mutect2 each used four threads per task, for a total of 16 threads across four concurrent tasks, with a maximum memory allocation of 8 GB in the variant calling step.
The choice of GATK as the analytical backbone reflects its standing in the field. The authors point out that GATK HaplotypeCaller and Mutect2 have performed well in comparative studies and are widely used in large projects such as The Cancer Genome Atlas, the UK Biobank whole-genome sequencing consortium and the 1000 Genomes Project. Mutect2 offers high sensitivity for detecting somatic variants even without a matched control, and the use of a panel of normals together with a germline resource helps filter false positives from its call sets. GATK’s probabilistic variant calling model, which considers base quality scores, mapping quality and variant quality scores, together with allele-specific filtering in FilterMutectCalls and machine-learning-based artifact filtering, is designed to distinguish true somatic variants from sequencing errors and germline variants. The authors also note that the recent GATK version 4.5.0.0 incorporates features aimed at functional equivalence with Illumina’s DRAGEN version 3.7.8, and that an implementation of the Smith-Waterman algorithm in HaplotypeCaller and Mutect2 significantly improves speed.
The motivation for building yet another pipeline lies partly in the cost of the alternatives. Commercial accelerated platforms such as Illumina DRAGEN and NVIDIA Clara Parabricks deliver enormous speed improvements but carry high licensing costs and require high-performance GPUs, putting them out of reach for smaller labs and groups with limited funds. Web platforms such as Galaxy offer accessible analysis but are constrained by limited storage and heavy global user demand, making joint variant calling from large sample sets difficult, and their workflows must be carefully designed to optimize parameter choices and minimize false positives and negatives. By contrast, WEAP runs on a user’s own hardware, integrates the latest recommended GATK protocols, automatically performs joint variant calls, false-positive filtering, genotyping and annotation, and, unlike many automated variant calling tools, supports somatic calling from both tumor-only and tumor-normal paired samples. The authors contrast this with the existing SeqMule automated workflow, which they say does not integrate the latest GATK best practices protocol.
The tool does have limitations, which the authors state plainly. WEAP does not support pausing or resuming an analysis once it is running, and it performs optimally on Linux, specifically Ubuntu 20.04 or higher, or on Windows 10 and 11 through the Windows Subsystem for Linux, on mid- to high-end computing configurations. Looking ahead, the team plans to add DeepVariant as an alternative variant caller, noting that a benchmark study found DeepVariant version 0.8.0 outperformed GATK version 4.1.2.0 on a trio sample, along with additional annotation tools, an implementation of the workflow in the Nextflow workflow manager, automated copy number analysis for tumor-normal paired samples and SNP calling from mitochondrial genomes. For now, WEAP is freely available on GitHub with its binary executables and a user manual, offering laboratories a route from raw sequencing data to annotated variant calls in a single automated step, at a fraction of the time required by serial processing.
Subject of Research: An automated open-source pipeline for accelerated germline and somatic variant calling in whole-exome sequencing data
Article Title: WEAP: an automated and accelerated pipeline for analyzing large scale whole exome sequencing data
Article References: Sarma, R. J., & Kumar, N. S. (2026). WEAP: an automated and accelerated pipeline for analyzing large scale whole exome sequencing data. BMC Methods, 3(1), Article 28. https://doi.org/10.1186/s44330-026-00080-9
Image Credits: AI Generated
DOI: 10.1186/s44330-026-00080-9
Keywords: whole-exome sequencing, WEAP, GATK, variant calling, bioinformatics pipeline, GNU parallel, somatic variants, germline variants, Mutect2, ANNOVAR, bwa-mem2, gastric cancer
News Source: Juliet Wilcox. (October 28, 2026). New Open-Source Pipeline Speeds Up Whole Exome Sequencing Analysis. Scienmag.
