Germline and Somatic Variant Calling

  • Germline and somatic variant calling are two major approaches used in bioinformatics to identify genetic alterations in biological samples. Both involve detecting differences between DNA sequences and a reference genome, but they differ in the biological origin of the variants, the samples analysed, the computational methods used, and their interpretation. Germline variants are inherited or arise in reproductive cells and may be transmitted to offspring, whereas somatic variants arise in non-reproductive cells during an individual’s lifetime and are generally restricted to particular tissues or cell populations. Understanding this distinction is essential in clinical genetics, cancer genomics, developmental biology, population genetics, and precision medicine.
  • Germline variants are genetic changes present in the inherited genome of an individual. They may be transmitted from parents through sperm or egg cells, or arise as new mutations in a parental germ cell or during early embryonic development. A germline variant is often present in most or all cells of the individual, although early developmental mosaicism can produce exceptions. Germline variants include single-nucleotide variants (SNVs), insertions and deletions (indels), copy-number variants (CNVs), and larger structural variants. They may be benign, harmful, or associated with increased susceptibility to particular diseases. Germline variant calling is commonly used to identify inherited genetic disorders, carrier status, disease-associated alleles, pharmacogenomic variants, and genetic differences among individuals or populations.
  • Somatic variants are genetic alterations acquired after fertilization in cells that are not part of the germline lineage. They arise through DNA replication errors, spontaneous chemical changes, exposure to radiation or chemicals, defects in DNA repair, and other biological processes. Somatic variants can occur in normal tissues, but they are particularly important in cancer because mutations may contribute to uncontrolled cell growth, tumour development, progression, and treatment resistance. Somatic variants can be restricted to a small group of cells or may become widespread within a tissue through clonal expansion. They are not normally inherited by offspring, although mutations arising in cells that contribute to reproduction can enter the germline and be transmitted.
  • The proportion of DNA molecules carrying a variant is often different in germline and somatic samples. In a typical diploid individual, a heterozygous germline SNV may have an alternate allele fraction near 50%, while a homozygous alternate variant may have an alternate allele fraction near 100%. These are approximate expectations rather than strict rules. Local copy-number changes, mapping bias, sequencing errors, contamination, aneuploidy, and mosaicism can alter the observed allele fraction. Somatic variants often occur at lower allele fractions because only a proportion of cells in the sample carry the mutation. In cancer samples, the observed variant allele fraction is influenced by tumour purity, subclonal composition, local copy number, loss of heterozygosity, and the presence of normal cells.
  • Germline variant calling generally begins with DNA extracted from blood, saliva, buccal cells, or another suitable tissue. Blood is frequently used in inherited-disease studies, but it may not always represent the individual’s complete genetic constitution. For example, blood can contain acquired clonal mutations, particularly in older individuals, and some mosaic variants may be absent or present at different levels in different tissues. DNA sequencing is performed using whole-genome sequencing (WGS), whole-exome sequencing (WES), targeted gene panels, or genotyping arrays, depending on the research or clinical objective. The sequencing reads are stored in FASTQ files and undergo quality control, adapter removal, and filtering before alignment to a reference genome.
  • Short-read germline variant discovery commonly uses alignment tools such as BWA-MEM2, followed by variant callers such as GATK HaplotypeCaller, DeepVariant, FreeBayes, or BCFtools. These tools identify candidate SNVs and small indels from aligned sequencing reads. GATK HaplotypeCaller uses local haplotype assembly and probabilistic genotype modelling, while DeepVariant uses a deep-learning approach to evaluate read evidence. FreeBayes applies a Bayesian framework, and BCFtools provides likelihood-based variant-calling methods. The selection of a caller depends on the sequencing platform, data quality, sample type, ploidy, and analysis requirements.
  • An illustrative germline variant-calling workflow using GATK HaplotypeCaller is:

# Generate a germline gVCF from aligned sequencing reads

gatk HaplotypeCaller \

    -R reference.fa \

    -I sample.sorted.bam \

    -O sample.g.vcf.gz \

    -ERC GVCF

  • The gVCF contains variant and reference-confidence information and is commonly used in multi-sample studies. When multiple individuals are analysed, their gVCFs can be combined or imported into a joint-genotyping workflow using tools such as GATK GenomicsDBImport and GenotypeGVCFs. Joint genotyping allows genotypes to be evaluated consistently across a cohort and can improve the representation of sites that are variable among samples.
  • An illustrative BCFtools germline calling command is:

# Call germline variants from an aligned BAM file

bcftools mpileup \

    -f reference.fa \

    -Ou sample.sorted.bam | \

bcftools call \

    -mv \

    -Oz \

    -o germline_variants.vcf.gz

# Index the resulting VCF

bcftools index germline_variants.vcf.gz

  • This example represents a simplified single-sample workflow. Actual analyses should account for sample ploidy, read groups, sequencing quality, mapping quality, variant representation, and the recommended settings for the selected caller. For non-diploid organisms, sex chromosomes, organellar genomes, and mixed-ploidy regions, appropriate ploidy settings may be required.
  • Somatic variant calling is more complex because the objective is to identify acquired mutations that may be present in only a fraction of the cells in a sample. In cancer genomics, a common approach is paired tumour-normal sequencing. DNA is extracted from tumour tissue and a matched normal sample, often from blood or another non-tumour tissue. Sequencing both samples allows the analysis to distinguish tumour-specific variants from inherited germline variants. Variants found in both samples are generally considered candidates for germline origin, while variants detected in the tumour but not in the matched normal sample are candidates for somatic origin. However, this distinction requires careful interpretation because sequencing errors, low coverage, tumour contamination of the normal sample, clonal haematopoiesis, mosaicism, and other biological factors can complicate classification.
  • The matched normal sample is especially important because common and rare germline variants may otherwise be incorrectly interpreted as tumour-acquired mutations. Population databases such as gnomAD and dbSNP can provide useful frequency information, but database absence does not prove that a variant is somatic. Conversely, a variant found in a population database may still be relevant to a particular tumour context. Matched normal sequencing provides direct sample-specific evidence and is generally preferable when the objective is to identify somatic mutations reliably.
  • Somatic variant calling commonly uses tools such as GATK Mutect2, Strelka2, VarDict, and other callers designed for tumour sequencing. Mutect2 is widely used for detecting somatic SNVs and small indels, particularly in paired tumour-normal workflows. It evaluates evidence from tumour and normal samples and can apply statistical models to distinguish candidate mutations from sequencing artifacts and germline variation. Strelka2 supports somatic and germline variant calling, while VarDict is used for variant detection in targeted and broader sequencing datasets. The choice of caller depends on tumour type, sequencing depth, sample design, variant classes, and the analytical requirements.
  • An illustrative paired tumour-normal somatic variant-calling command using GATK Mutect2 is:

# Call candidate somatic variants using tumour and matched normal samples

gatk Mutect2 \

    -R reference.fa \

    -I tumour.sorted.bam \

    -I normal.sorted.bam \

    -normal NORMAL_SAMPLE \

    -O somatic_candidates.vcf.gz

  • The sample name supplied with -normal must match the normal sample identifier in the alignment file’s read-group information. This command produces candidate calls; it is not a complete somatic analysis. A typical Mutect2 workflow also includes contamination estimation, orientation-bias assessment where appropriate, and filtering using tools such as FilterMutectCalls. The exact steps and resources depend on the sample type and GATK version.
  • Somatic variant filtering requires particular attention because tumour samples may contain mutations at low variant allele fractions. Sequencing errors, oxidative damage, formalin-related artifacts in fixed tissue, strand bias, mapping problems, and low-complexity regions can produce false-positive calls. Filtering methods evaluate evidence such as read depth, base quality, mapping quality, allele fraction, strand orientation, read position, and support in the matched normal sample. Tumour purity and contamination estimates are also important. A mutation may be biologically real but difficult to detect when it is present in only a small fraction of tumour cells or when the sequencing depth is insufficient.
  • A simplified Mutect2 filtering example is:

# Filter candidate somatic variants

gatk FilterMutectCalls \

    -R reference.fa \

    -V somatic_candidates.vcf.gz \

    -O somatic_filtered.vcf.gz

  • Somatic workflows may also estimate contamination using tools such as GATK GetPileupSummaries and CalculateContamination. These steps help identify contamination from normal cells or other sources that can affect somatic allele-fraction estimates. Additional artifact filtering may be required for particular sequencing platforms, library preparations, or tissue types.
  • The variant allele fraction (VAF) is an important measure in both germline and somatic variant analysis. It represents the proportion of observed sequencing reads at a locus that support the alternate allele, although the precise calculation may depend on the caller and available read evidence. A heterozygous germline variant in a typical diploid region often has a VAF near 0.5. A somatic mutation may have a lower VAF because it is present in only a subset of tumour cells. However, VAF alone cannot reliably distinguish germline from somatic variants. Tumour purity, copy-number alterations, loss of heterozygosity, subclonal structure, and normal-cell admixture can all change the expected VAF. For this reason, paired tumour-normal analysis and additional biological evidence are essential.
  • Mosaicism is an important situation that complicates the distinction between germline and somatic variation. A mosaic individual carries two or more genetically distinct cell populations that originated from a single fertilized egg. A variant arising during early embryonic development may be present in several tissues but not in every cell. A post-zygotic mutation arising later may be restricted to a particular tissue or organ. Mosaic variants can be relevant to developmental disorders, neurological conditions, cancer predisposition, and congenital abnormalities. Their detection may require high-depth sequencing, analysis of more than one tissue, and specialised variant-calling methods. A variant detected in blood at a low allele fraction may represent constitutional mosaicism, an acquired blood-cell mutation, or a technical artifact; interpretation requires clinical and laboratory context.
  • Clonal haematopoiesis is another important consideration, particularly when blood is used as the matched normal sample in cancer studies. In this condition, acquired mutations arise in blood-forming stem or progenitor cells and become detectable in a proportion of blood cells. Such mutations can appear in blood-derived normal DNA even though they are not inherited constitutional variants. If the same mutation is present in a tumour sample, it may be incorrectly classified unless the possibility of clonal haematopoiesis is considered. Additional tissue testing and careful evaluation of genes commonly associated with clonal haematopoiesis may help resolve uncertain cases.
  • Long-read sequencing is increasingly used for both germline and somatic variant analysis. PacBio HiFi and Oxford Nanopore technologies generate reads that can span repetitive sequences, complex rearrangements, and larger structural variants. Long reads can improve the detection of insertions, deletions, repeat expansions, inversions, translocations, and complex genomic events. They can also provide useful information about haplotypes and the relationship between nearby variants. For germline analysis, long-read callers such as DeepVariant and Clair3 may be used with supported data types and workflows. For somatic structural variant analysis, tools such as Sniffles2 and other long-read structural variant callers can identify candidate rearrangements from tumour sequencing data. The suitability of each tool depends on the sequencing platform, read accuracy, coverage, and variant class.
  • Short-read sequencing remains widely used for germline and somatic variant discovery because it offers high throughput and well-established analytical pipelines. Short reads are generally effective for SNVs and small indels in regions that can be mapped reliably. Long reads offer advantages in complex genomic regions and structural variant detection, but performance depends on read accuracy and coverage. Combining short-read and long-read data can provide complementary evidence, particularly when a study requires reliable small-variant detection together with improved structural variant and haplotype resolution.
  • Structural variant analysis is relevant to both germline and somatic studies. Germline structural variants may contribute to inherited disorders, developmental conditions, and variation in complex traits. Somatic structural variants can drive cancer through gene fusions, oncogene amplification, tumour-suppressor gene disruption, and chromosomal rearrangements. Short-read callers such as Manta, DELLY, and GRIDSS use paired-end and split-read evidence, while long-read callers such as Sniffles2 and cuteSV can use long alignments that span breakpoints. Copy-number analysis may require additional tools, including CNVkit, GATK gCNV, or other methods selected for the sequencing design. Structural variant interpretation should consider breakpoint uncertainty, supporting reads, copy-number context, and possible complex rearrangements.
  • After variant calling and filtering, variants are commonly annotated to determine their possible biological and clinical significance. Annotation tools include Ensembl Variant Effect Predictor (VEP), SnpEff, and ANNOVAR. These tools can identify consequences such as synonymous, missense, nonsense, frameshift, splice-site, and regulatory-region variants. Germline variant interpretation may use resources such as ClinVar, ClinGen, gnomAD, OMIM, and the Human Phenotype Ontology. Somatic cancer variant interpretation may use resources such as COSMIC, OncoKB, CIViC, and relevant clinical guidelines. Database annotations should be treated as evidence to evaluate rather than definitive conclusions. Clinical significance depends on the strength of evidence, disease context, inheritance pattern, tumour type, and the quality and relevance of the underlying data.
  • Germline and somatic variants have different implications for genetic counselling and clinical care. Germline variants may indicate inherited disease risk, carrier status, or susceptibility to future conditions. They can also have implications for biological relatives and reproductive decisions. Somatic variants are generally relevant to the tissue or tumour in which they arise and may inform diagnosis, prognosis, treatment selection, and monitoring. In some circumstances, a tumour analysis identifies a variant that may also be germline. Such findings require appropriate confirmatory testing using a suitable non-tumour tissue and, where relevant, clinical genetics assessment. Tumour sequencing alone should not be used to establish inherited status.
  • Variant calling performance is evaluated using measures such as precision, recall, and F1 score. Precision describes the proportion of called variants that are true positives, while recall describes the proportion of true variants detected by the pipeline. Benchmarking resources such as Genome in a Bottle (GIAB) are widely used for evaluating human germline SNV and small-indel calling in well-characterised regions. Somatic benchmarking is more challenging because tumour samples can contain subclonal mutations, variable purity, and complex copy-number changes. Reference materials and synthetic mixtures can help evaluate somatic callers at different allele fractions. Tools such as hap.py and benchmarking frameworks based on truth sets can be used for small-variant comparisons, while structural variant benchmarking may use tools such as Truvari. Benchmark results must be interpreted in light of the genomic regions, variant classes, and sample characteristics represented by the truth set.
  • Several limitations affect germline and somatic variant calling. Low sequencing depth can cause false negatives, while sequencing errors and alignment artifacts can produce false positives. Repetitive regions and duplicated genes can make read mapping ambiguous. Reference genome bias may reduce the detection of alleles that differ substantially from the reference. Somatic analysis is particularly affected by tumour heterogeneity, low tumour purity, subclonal mutations, normal-cell contamination, and variable sequencing depth. Germline analysis can be complicated by mosaicism, unusual ploidy, consanguinity, and acquired mutations in the tissue used for testing. No variant caller detects every variant class equally well, and results should be validated when they have important clinical or biological consequences.
  • Reproducibility is essential in germline and somatic variant analysis. Researchers should document the reference genome assembly and version, sequencing platform, sample type, library preparation, quality-control procedures, alignment software, variant caller and version, parameters, filtering criteria, normalization steps, and annotation databases. Germline studies should record relevant family relationships and inheritance assumptions where appropriate. Somatic studies should document tumour type, tumour purity, matched-normal availability, sequencing depth, and contamination estimates. Workflow systems such as Nextflow, Snakemake, and Cromwell can help automate analyses and maintain consistent records. Software environments can be managed using Conda or containers, while logs and intermediate files support auditing and troubleshooting. Human genomic data also require appropriate privacy, consent, and access controls.
  • Germline and somatic variant calling are essential but distinct components of modern genomic analysis. Germline calling focuses on inherited and constitutional genetic variation, while somatic calling identifies acquired mutations that may be restricted to particular tissues or cell populations. Germline workflows commonly use single-sample or cohort-based analysis, whereas somatic cancer workflows often rely on paired tumour-normal sequencing and specialised filtering. Short-read and long-read technologies provide complementary evidence for detecting SNVs, indels, and structural variants. Accurate interpretation requires appropriate algorithms, careful quality assessment, biological context, validation, and reproducible computational procedures.
Author:

Leave a Reply

Your email address will not be published. Required fields are marked *