Genome variant calling

From reads to a short list of variants actually worth looking at — not a file with a million rows.

When variant calling is the right tool

data2biology detects genomic variants — SNPs, short indels, structural variants and copy number changes — from whole-genome, exome and targeted panel data, in both germline and somatic modes. The analyses are for research purposes and do not replace genetic testing performed by a diagnostic laboratory. Variant calling answers the question of how an individual, strain or line differs from a reference sequence. It is used in population genetics, breeding and marker-assisted selection, disease research, cell line characterisation, and analyses of somatic changes in tumour material.

We support whole-genome, exome and targeted panel sequencing, as well as long-read data, which handles structural variants and repetitive regions better. We run both germline and somatic analyses, the latter using paired tumour–normal samples.

Our analyses are for research purposes. They are not an in vitro diagnostic medical device and do not replace genetic testing performed by a diagnostic laboratory or clinical interpretation by a medical geneticist.

How the data flows

Below we describe how we typically run this type of analysis. It is not a fixed procedure: the choice of tools depends on the organism, library type, sequencing depth and research question, and standards in bioinformatics change quickly. We agree the methodology with you before the project starts and update it as tools and reference databases evolve.

Input
FASTQ files · reference genome · a BED file for exome or panel designs · paired tumour–normal samples for somatic analysis
Read quality control and preparation
Quality assessment plus removal of adapters and low-quality reads before mapping.
FastQCfastpMultiQC
Mapping to the reference genome
The algorithm depends on read length. We agree the genome build with you, since it determines compatibility with annotation databases.
BWA-MEM2minimap2samtools
Alignment post-processing
Sorting, marking of PCR duplicates and base quality recalibration where suitable databases exist for the species.
GATK MarkDuplicatessamtoolsGATK BQSR
Coverage assessment
Depth distribution and coverage uniformity. For panels and exomes we identify which target regions fell below the threshold for reliable variant calling.
mosdepthQualimapPicard
Short variant calling
Detection of SNPs and short indels. Germline and somatic variants use different statistical models, because their expected allele fractions differ.
GATK HaplotypeCallerDeepVariantMutect2Strelka2bcftools
Variant filtering and QC
Removal of artefactual calls using a quality model or threshold filters, and in somatic analyses additionally with a panel of normals.
GATK VQSRbcftools filterpanel of normals
Structural variants and copy number changes
Deletions, duplications, inversions and translocations are detected with dedicated tools matched to the data type. Sensitivity is markedly higher with long reads.
MantaDellySniffles2CNVkit
Variant annotation
Assignment of gene context, predicted protein-level consequence and frequency in population reference databases.
VEPSnpEffgnomADClinVardbSNP
Prioritisation and reporting
Narrowing the list by agreed criteria: variant type, population frequency, membership of a candidate gene list, mode of inheritance.
RPythonR Markdown
Output
Raw, filtered and annotated VCF files · variant table in Excel · structural variants and CNVs · coverage report · BAM/CRAM files
A typical data flow for variant calling. Steps 5–7 are matched to the analysis mode: germline or somatic.

What we pay attention to

Coverage decides what you will not see

A missing variant may mean it is absent, or that coverage at that position was too low to say anything. This distinction is often skipped, yet it is central to interpreting a negative result. Our report therefore always states what fraction of target regions reached the thresholds for reliable calling, and which specific genes fell below.

Filtering is a trade-off, not absolute truth

Any filter set shifts the balance between false positives and the risk of discarding a true variant. When screening candidate genes, looser filtering usually pays off; in genome-wide analysis, stricter. We agree this at the outset and deliver both the filtered and the complete version, so a discarded variant can always be revisited.

Pathogenicity predictions are hints

Tools predicting the effect of an amino acid change rely on evolutionary conservation and structural models. Their agreement with reality is limited and varies between genes. We report these values as one input alongside population frequency, variant type and published evidence, never as a verdict on their own.

What you receive

File Format What it contains
variants.filtered.vcf.gz + .tbi VCF The standard variant exchange format, compressed and indexed. One row per position, readable by all common tools and browsers.
  • CHROM, POS, REF, ALT — position plus reference and alternative allele
  • QUAL, FILTER — call quality and whether the variant passed filtering
  • INFO — annotations: gene, consequence, population frequency, depth
  • Sample columns with genotype (GT), depth (DP), allele depth (AD) and genotype quality (GQ)
variants.raw.vcf.gz VCF The complete call set before filtering — it lets you revisit a variant that filtering removed.
variants_annotated.xlsx XLSX The same information as a spreadsheet, ordered by priority. The file you will actually work with day to day.
  • Chromosome, position, alleles, genotype and zygosity
  • Gene, transcript and the change in HGVS notation at nucleotide and protein level
  • Variant type and predicted severity: missense, nonsense, frameshift, splice site
  • Frequency in population reference databases and entries from clinical databases
  • Coverage depth and alternative allele fraction in the sample
  • Separate sheets: rare variants, variants in candidate genes, analysis parameters
structural_variants.vcf / cnv.xlsx VCF + XLSX Structural variants and copy number changes, where in scope.
  • Change type, start and end coordinates, length
  • Number of supporting reads and a confidence measure
  • Genes affected, with a note on whether the change is full or partial
coverage_report.html HTML The coverage report — the file that determines how a negative result should be read.
  • Mean and median depth for each sample
  • Fraction of target regions covered above successive thresholds
  • A list of regions and genes with coverage insufficient for analysis
alignments/*.cram + *.crai CRAM Aligned reads in a space-efficient format, with indexes. They let you inspect every reported variant yourself in IGV.
methods.docx DOCX A methods section draft with tool versions, the reference genome build, the filters applied and citations.

What we need from you

  • Raw FASTQ files or existing BAM/CRAM alignments, with the genome build used
  • A BED file with target regions, for exome or targeted panel designs
  • Whether the analysis is germline or somatic, and for the latter, paired samples
  • A candidate gene list or region of interest, where one exists
  • For human data: the legal basis for processing and participant consents

Human genetic data is a special category of personal data. We take on analyses of such data after the processing arrangements have been agreed in advance — please contact us before sending it.

Quotation and how we work

We quote every project individually. The cost depends mainly on the number of samples, the availability of a reference genome, the number of comparisons and the scope of analyses you ask for.

  1. You tell us what you want to find out and what data you have.
  2. If anything needs clarifying we arrange a short call — free of charge and without obligation.
  3. Within 3–5 working days you receive a quote with the scope of work, the list of result files and a delivery date.
  4. You decide. The quote carries no obligation.

The scope and cost of projects like these vary widely enough that quoting a single representative figure would be misleading. If you need a ballpark before a formal quote, just ask — we can give a rough range straight away.

Any figures given are indicative and net of tax; VAT is added according to the applicable regulations. They do not constitute a binding offer. More about how we work in the FAQ.

Got data to analyse?

Tell us what you want to find out and we will scope the analysis and timeline together. Discussing the project and preparing a quote are free of charge.

Request a quote See a worked example