Genomics & Variant Analysis
Germline SNVs and indels from BAMs, gVCFs, or scoped FASTQs via pinned gVCF-to-VCF workflows, figures, scripts, and a Methods draft [1].
| Category | Details |
|---|---|
| Supported platforms / instruments | Illumina NovaSeq X / 6000 / NextSeq 2000, HiSeq 2500/4000; MGI DNBSEQ-T7 / G400 / MGISEQ-2000 (BGI) short-read BAMs; Element Biosciences AVITI and Ultima Genomics UG100 when scoped at kickoff. Input: coordinate-sorted .bam/.cram, per-sample .g.vcf.gz, or paired-end FASTQ when upstream alignment is in scope |
| Input requirements | WGS BAMs: ≥30× mean autosomal depth and >95% callability recommended [9]. WES/panel BAMs: ~100× mean on-target depth typical [10]. Cohort VQSR: typically ~30 exomes or one high-quality WGS per GATK guidance [2]. Valid @RG read groups required |
| Reference builds supported | Human GRCh38 primary (GATK resource bundle 4.6); legacy GRCh37/hg19 on request; mouse GRCm39; custom references scoped at kickoff |
| Primary tools (with versions) | GATK 4.6.0.0; DeepVariant 1.8.0; Strelka2 2.9.10 (optional); Picard 3.2.0; samtools 1.21; bcftools 1.21; mosdepth 0.3.3; Ensembl VEP 112; BWA-MEM2 2.2.1 when alignment is in scope; MultiQC 1.25 |
| Typical turnaround time | 1-3 weeks (single-sample from BAM); 3-6 weeks (multi-sample cohort with joint genotyping and VQSR). Confirmed at kickoff |
| Deliverable formats | .g.vcf.gz, joint and filtered .vcf.gz, variant_annotation_master.tsv, QC tables; PDF/SVG figures; HTML QC report; documented bash/R/Python scripts; Methods draft |
| Key cited best-practice reference | Van der Auwera et al. [1], Current Protocols in Bioinformatics; Van der Auwera & O'Connor (2020), Genomics in the Cloud; Barbitoff et al. [11], BMC Genomics |
| Custom / bespoke analysis | Non-standard inputs, outputs, and methods scoped at kickoff, e.g. re-calling vendor VCFs, multi-caller benchmarks, GLnexus merge for DeepVariant cohorts, pedigree filtering, plink exports, or integration with existing gVCFs |
Variant calling identifies SNVs and short indels from aligned reads, not library prep or basecalling. GATK HaplotypeCaller assembles local haplotypes and assigns genotypes; joint genotyping can rescue variants that are weakly supported in a single sample [1]. Unlike CNV/SV calling at segment scale, SNV/indel calling works at single-base resolution. Documented pipelines pin software versions and filter logic so the analysis can be reproduced [3, 4]. GATK processed ~180,000 WGS samples in TOPMed; DeepVariant was selected for 500,000+ UK Biobank exomes [5]. Pepkio starts from client BAMs, gVCFs, or scoped FASTQs and returns filtered, annotated VCFs. Custom scope is agreed at kickoff. See the variant calling glossary.
Dedicated variant calling fits when alignments already exist, vendor VCFs need re-calling, or a cohort needs joint genotyping that is consistent across batches.
| Approach | Best for | Limitations | Approximate cost range |
|---|---|---|---|
| Dedicated variant calling (BAM/gVCF in) | Re-calling, cohort harmonization, filter tuning, multi-caller benchmarks, plink or GWAS-ready exports | Requires an upstream BAM QC audit; does not recover failed library prep; somatic tumor-normal work is scoped separately | Bioinformatics-only; usually lower than full FASTQ-to-VCF when alignment already exists |
| Full WGS/WES spoke | End-to-end FASTQ-to-VCF with modality-specific alignment and depth QC | Larger scope when BAMs already exist | Library prep + sequencing + bioinformatics vary by depth and sample count |
| Vendor or sequencer default VCF | Fast turnaround bundled with sequencing | Opaque filters, version drift, and weaker cross-center harmonization [3, 4] | Often bundled with sequencing; re-analysis is needed only if the calls are not adequate |
Pepkio verifies BAM, CRAM, gVCF, or FASTQ integrity, confirms coordinate sorting and indexing, and checks @RG read groups required by GATK [1]. Sample sex, batch, sequencing center, and reference build are recorded in sample_manifest.csv; reference-build mismatches are flagged before calling.
Tools used: md5sum; samtools quickcheck; custom validation scripts
Output: sample_manifest.csv with sample IDs, input type, reference build, read counts or BAM size, and QC flags
Pepkio inspects duplicate marking and base quality score recalibration (BQSR) status. Calling without MarkDuplicates or ApplyBQSR can increase false-positive SNVs and indels [1, 12]. Missing prerequisites trigger scoped Picard MarkDuplicates and GATK BQSR re-runs, or documented deviation notes.
Tools used: Picard 3.2.0 CollectAlignmentSummaryMetrics; GATK 4.6.0.0 ValidateSamFile; custom BAM audit scripts
Output: bam_prerequisite_audit.csv with duplicate-marking status, BQSR tag presence, mapping rate, and re-preprocessing recommendations
mosdepth and Picard CollectWgsMetrics or CollectHsMetrics report mean depth, breadth at ≥15×/≥20×, and on-target fraction [9, 12]. Callable breadth, not mean depth alone, determines locus-level calling; SNV reproducibility plateaus above ~30×, while indel reproducibility continues to improve with depth [4]. Sub-threshold samples are flagged before discovery.
Tools used: mosdepth 0.3.3; Picard 3.2.0 CollectWgsMetrics or CollectHsMetrics
Output: coverage_summary.csv; mosdepth.global.dist.txt or target-coverage tables; coverage histogram and CDF plots
Intervals are set to whole genome, client exome or panel BED, or agreed subsets. GATK resource bundle 4.6 known sites for BQSR verification and VQSR training match the reference build [2]. Interval lists and BED files are version-controlled in the deliverable bundle.
Tools used: GATK 4.6.0.0 PreprocessIntervals (when needed); custom interval validation scripts
Output: calling_intervals.list or .bed; interval_validation_report.txt
GATK HaplotypeCaller runs in -ERC GVCF mode, emitting gVCF blocks across uncalled regions [1]. DeepVariant 1.8.0 or Strelka2 2.9.10 are optional when scoped at kickoff [11, 13, 14]. Per-sample Ti/Tv ratios are checked against ~2.0-2.1 for whole-genome germline SNVs [12].
Tools used: GATK 4.6.0.0 HaplotypeCaller; DeepVariant 1.8.0 or Strelka2 2.9.10 when scoped
Output: {sample}.g.vcf.gz and .tbi; per-sample variant count and Ti/Tv summary
For multi-sample projects, gVCFs import into GenomicsDB and joint-genotype with GenotypeGVCFs [1]. Single-sample projects skip this step. GLnexus merge is scoped for DeepVariant cohorts when required [3].
Tools used: GATK 4.6.0.0 GenomicsDBImport; GATK GenotypeGVCFs; GLnexus when scoped for DeepVariant
Output: {cohort}.joint.vcf.gz; GenomicsDB workspace; sample count and site-level summary
When cohort size supports it, GATK VariantRecalibrator applies VQSR trained on known polymorphism resources [2]. Smaller cohorts use documented GATK hard filters with all FILTER labels retained [12]. DeepVariant projects use quality thresholds documented in the Methods draft. Filtered and pass-only VCFs export separately.
Tools used: GATK 4.6.0.0 VariantRecalibrator / ApplyVQSR or VariantFiltration; bcftools 1.21
Output: {cohort}.filtered.vcf.gz; {cohort}.pass-only.vcf.gz; VQSR tranche plots or hard-filter summary table
Pepkio computes Ti/Tv on pass sites, variant counts by type, and FILTER distributions. In a multi-center WGS benchmark, caller and aligner choice affected reproducibility more than sequencing platform [4]; multi-caller benchmarks are documented when scoped. Problematic regions such as segmental duplications and homopolymers are flagged.
Tools used: bcftools 1.21 stats; custom QC scripts; hap.py when benchmark comparison is scoped
Output: variant_qc_summary.csv; sample_qc_summary.csv; filter_summary.csv; Ti/Tv plot; FILTER status distribution chart; optional benchmark comparison table
Ensembl VEP annotates consequence terms, gene symbols, and population or clinical fields when databases are configured [15]. ClinVar, gnomAD, or client gene lists are added when scoped at kickoff.
Tools used: Ensembl VEP 112; bcftools 1.21
Output: variant_annotation_master.tsv; annotated .vcf.gz when requested
MultiQC aggregates QC metrics when multiple stages are in scope [16]. Scripts, README, Methods draft with exact versions and filter parameters, and HTML QC report are packaged per agreed retention policy.
Tools used: MultiQC 1.25; custom packaging scripts
Output: deliverable bundle with scripts, figures, QC report, and Methods draft
For WGS BAMs, Pepkio recommends ≥30× mean autosomal depth with >95% callability [9]. WES or panel BAMs typically need ~100× mean on-target depth [10]. VQSR requires at least one high-quality WGS or ~30 exomes [2]; smaller cohorts use hard filters. Thresholds are confirmed at kickoff.
Yes. This is the default entry point. Pepkio audits duplicate marking, BQSR status, read groups, and reference build before calling [1]. Compatible client .g.vcf.gz files can enter at joint genotyping. Missing BAM prerequisites can be re-processed when scoped at kickoff.
Yes, with caveats. Mean depth below ~20× for WGS or ~50× on-target for exome tends to reduce sensitivity, especially for indels >5 bp [4]. Sub-threshold samples are flagged in sample_qc_summary.csv. Partial calling on high-confidence intervals is possible when re-sequencing is not feasible.
Illumina and MGI DNBSEQ short-read BAMs use the standard GATK or DeepVariant workflow on coordinate-sorted alignments. Element AVITI and Ultima UG100 BAMs are supported when scoped at kickoff with validated aligner compatibility. PacBio and Oxford Nanopore SNV/indel calling is referred to the long-read DNA-seq spoke.
Single-sample projects from client BAMs typically complete in 1-3 weeks. Multi-sample cohorts with joint genotyping and VQSR typically take 3-6 weeks. Timelines depend on sample count, upstream BQSR needs, and annotation scope, and are confirmed at kickoff.
Harmonized pipelines aligned to functional-equivalence standards can reduce batch-driven differences [3]. Pepkio stratifies sequencing center, flowcell, and batch in QC reports. Joint genotyping can improve sensitivity for weakly supported variants across batches [1]. Cross-batch harmonization beyond standard QC is scoped at kickoff when needed.
Yes. You retain full ownership of code, scripts, and results. Pepkio delivers commented bash, R, and Python scripts with environment lock files. Jupyter or R Markdown delivery is available on request.
Yes. Checkpoint reviews occur after BAM prerequisite audit, coverage assessment, and before final delivery. A PhD-level scientific contact leads the project at agreed milestones.
Clarification of methods, filter logic, QC thresholds, and minor figure or table revisions within agreed scope (typically ≤20% of deliverables). Substantial new reviewer requests are scoped separately.
No. Pepkio operates as a fee-for-service provider and does not require co-authorship unless that is discussed in advance.
GATK HaplotypeCaller gVCF is the default for multi-sample joint genotyping and VQSR [1]. In a systematic GIAB WES and WGS benchmark, DeepVariant showed the best performance and highest robustness and is preferred for single-sample accuracy when scoped [11, 13]. Multi-caller benchmarks against vendor VCFs are scoped at kickoff when discordance is a concern [4].
Yes. Bespoke work such as multi-caller consensus, re-calling vendor VCFs, tumor-normal somatic extensions, pedigree filtering, plink exports, custom annotation, or gVCF integration is scoped at kickoff with milestone pricing [12].
Tell us:
We'll tell you:
Contact us to start with a free consultation. Need everyday bench calculators? Try our free lab tools.
