Somatic Mosaic Variant Calling

Somatic mosaic variant calling identifies genetic variants that arise postzygotically and are present in only a subset of cells. Unlike germline variants inherited and present in every cell, mosaic variants occur during development or aging and exist at variable allele fractions, often below 20 percent. This guide is for researchers and bioinformaticians who need to detect low allele fraction somatic variants in normal tissues, aging studies, or noncancerous disease contexts. Use this framework to understand core concepts, select appropriate tools, avoid common pitfalls, and interpret results with appropriate caution. The Landscape of parental postzygotic mutations across more than 11,000 rare disease trios illustrates the clinical relevance of mosaic variants in apparently healthy parents. Meanwhile, a comprehensive review on somatic variant calling beyond cancer explains how algorithms originally designed for tumor analysis can be repurposed for low allele fraction detection in any genomic study.

At a Glance

Decision Criteria for Using Somatic Mosaic Calling

Use somatic mosaic variant calling when you expect variants that are absent from germline DNA or present at much lower frequencies than heterozygous germline calls. This applies to studies of aging, where somatic mutations accumulate in blood and tissues as shown by the mechanism of age related accumulation of mtDNA mutations in human blood. It also applies to rare disease trios where a de novo mosaic variant may explain a phenotype despite being absent from most reads. Do not use standard germline calling pipelines for mosaic detection because they assume a 50 percent or 100 percent allele fraction model. Instead, incorporate tumor normal calling algorithms that model variant allele fractions continuously.

Key decision points include:

• Expected variant allele fraction (VAF). If VAF is above 20 percent, standard cancer somatic callers with matched normal may suffice. For VAF between 1 and 20 percent, you need specialized tuning and careful filtering.
• Sequencing depth. Reliable detection of a 5 percent mosaic variant requires at least 100x depth. At 1 percent VAF, depths of 500x or more are recommended. Consult resources like the Galaxy Training Network for depth recommendations in somatic variant analysis.
• Matched control. A matched normal tissue or a known germline sample is essential to filter inherited variants. Without it, you cannot distinguish a low frequency germline variant from a true mosaic somatic event.
• Tissue source. Different tissues have different background mutation rates. Somatic variant detection in normal tissues from single cell sequencing data reveals that even healthy organs harbor many low frequency variants (Somatic variant detection in normal tissues from single cell sequencing data).

Practical Workflow for Somatic Mosaic Variant Calling

The workflow below is a practical sequence. Adapt it to your data type (WES, WGS, targeted panel, or mtDNA sequencing).

1. Read preprocessing. Trim adapters, remove low quality bases, and align to a reference genome using BWA MEM or similar. For mitochondrial mosaic analysis, use a dedicated aligner that accommodates the circular genome. The NCBI Sequence Read Archive provides raw data for testing pipeline parameters.

2. Base quality score recalibration (BQSR). Use GATK or a recalibration tool to correct systematic base quality errors. This step disproportionately reduces false positives at low VAF.

3. Run a somatic caller with matched normal. Tools like Mutect2 (part of GATK) or Strelka2 require a tumor bam and a matched normal bam. Set the “tumor” as your mosaic sample and the “normal” as a germline control (often blood or buccal). Use the caller’s parameter for minimum VAF. For Mutect2, set --minimum-allele-fraction 0.01 if you want to capture 1 percent variants.

4. Filter artifacts using built in filters. Somatic callers apply statistical models to remove sequencing errors. However, these filters are calibrated for cancer samples with higher VAF. Reduce filter stringency stepwise and validate with orthogonal methods. The Bioconductor package VariantTools can help with post calling filtration and annotation.

5. Annotate and prioritize. Use tools like ANNOVAR or VEP to add gene context, population frequency, and predicted functional impact. Remove any variant present in gnomAD at a population frequency above 0.1 percent, as it is likely a rare germline variant rather than a true mosaic.

6. Validate with orthogonal sequencing. A second sequencing technology (e.g., amplicon based ultra deep sequencing) confirms low frequency variants. For mtDNA mosaic detection, long read sequencing improves resolution of heteroplasmy as shown by Long read MitoScope.

Quality Checks

Evaluate these metrics to assess the reliability of your mosaic calls:

• Strand bias. True mosaic variants should appear on both forward and reverse strands. Use Fisher’s exact test to remove calls with significant strand bias.
• Mean base quality of alternate allele. Low quality calls at the variant site indicate false positives. Exclude variants where mean base quality for the alternate allele is less than Q20.
• Read position bias. Artifacts often cluster at read ends. Check that supporting reads are distributed across the full read length.
• Normal contamination. Ensure the matched normal sample has fewer than 2 reads supporting the variant. If present, the variant might be a low level germline mosaic present in the control tissue.

The EMBL EBI Training offers modules on quality control for variant calling, including metrics specific to somatic detection.

Common Mistakes

• Using germline callers without adjustment. Germline callers assume binomially distributed allele counts with expected VAF of 0.5 or 1.0. Calling mosaic variants with these tools produces excessive false negatives. Always use a caller that models a continuous VAF spectrum.
• Ignoring the matched normal requirement. Without a matched normal, you cannot differentiate inherited rare variants from true mosaics. The result is an inflated list of candidate variants, most of which are false positives.
• Setting VAF thresholds too high. Many cancer somatic callers set a default minimum VAF of 10 percent. For mosaicism in normal tissue or aging studies, you need to lower it to 1 percent or 2 percent. Failing to adjust this parameter misses the majority of mosaic events.
• Overlooking tissue specific mosaicism. Mosaic variants in one tissue may not be present in the matched blood normal. If you use blood as the normal control but the mosaic is in brain, the variant will appear as somatic but may actually be a germline variant in blood. This is especially relevant for studies using FFPE tumor tissue as described in prostate cancer genetics in India.
• Not validating with orthogonal methods. Calling a mosaic variant from a single sequencing run is risky. Validate with an independent method such as digital droplet PCR or targeted resequencing at 10,000x depth.

Limits and Uncertainty

Somatic mosaic variant detection has fundamental limits. The most important is the trade off between sensitivity and specificity. At very low allele fractions (below 1 percent), sequencing error rates become comparable to true variant frequencies. Even with careful filtering, false positive rates can exceed 50 percent. The Mutat Research Review emphasizes that below 0.5 percent VAF, detection becomes unreliable without extremely deep sequencing (1000x or more).

Tissue sampling introduces another uncertainty. A mosaic variant present in only a few cells may be missed if the sampled tissue does not include those cells. This is why single cell sequencing offers more resolution but at higher cost. The bioRxiv preprint on single cell somatic variant detection shows that bulk sequencing averages out cellular heterogeneity.

Interpretation of clinical significance remains challenging. A mosaic variant in a disease gene with a VAF of 2 percent may or may not contribute to phenotype. Functional studies or testing in affected tissues are often required. Additionally, mitochondrial mosaicism accumulates with age and may be a benign marker rather than a disease driver (age related mtDNA accumulation).

Finally, tools are not standardized. No single pipeline works for all data types. The NCBI Bookshelf offers technical reference on variant calling best practices but does not endorse a universal method.

Frequently Asked Questions

1. What allele fraction threshold should I use for calling mosaic variants?

The threshold depends on your sequencing depth and tolerance for false positives. At 100x depth, a minimum VAF of 5 percent is practical. At 500x depth, you can reliably call down to 1 percent. For research studies, a threshold of 2 percent with validation is common. For diagnostic applications, a more conservative 10 percent threshold is often used.

2. Can I call mosaic variants without a matched normal sample?

You can, but the false positive rate will be very high. Without a matched normal, you cannot distinguish inherited low frequency variants from somatic ones. If a matched normal is unavailable, use population frequency filters and careful manual review, but acknowledge the high uncertainty.

3. How do mosaic variants in mitochondria differ from nuclear mosaic variants?

Mitochondrial DNA is present in many copies per cell, so mosaic variants (heteroplasmy) appear at any fraction from 0 to 100 percent. Calling mtDNA mosaicism requires specialized tools that account for the circular genome and high copy number. The Long read MitoScope preprint demonstrates a method that separates true mitochondrial variation from nuclear mitochondrial sequences (NUMTs).

4. Why do I get more somatic mosaic variants than expected?

High numbers likely come from technical artifacts, poor base quality, or contamination. Check your normal sample for evidence of the same variants at low level. Also verify that you have applied proper filtering for strand bias and read position. If the number remains high, reduce the VAF threshold or require support from both directions.

References and Further Reading

• NCBI Bookshelf. Variant Calling Best Practices. General reference for genomic analysis.
• EMBL EBI Training. Quality Control in Sequencing Data. Practical modules on QC metrics.
• Galaxy Training Network. Somatic Variant Calling Workflows. Step by step tutorials for Galaxy.
• Bioconductor. VariantTools Package. Software for variant filtering and annotation.
• NCBI Sequence Read Archive. Public Repositories for Sequencing Data. Source of raw data for testing pipelines.
• Landscape of parental postzygotic mutations across more than 11,000 rare disease trios. Am J Hum Genet. PubMed 42442367.
• Somatic variant calling beyond cancer: Repurposing algorithms to map low allele fraction variants across genomics. Mutat Res Rev Mutat Res. PubMed 42372426.
• Somatic variant detection in normal tissues from single cell sequencing data. bioRxiv. PubMed 42327286.
• Mechanism of age related accumulation of mtDNA mutations in human blood. Nature. PubMed 42203880.
• Toward the genetic landscape of prostate cancer in India: whole exome and low pass whole genome sequencing of FFPE tumors. Front Syst Biol. PubMed 42246034.

Related Articles

• Structural Variant Validation
• Repeat Expansion Sequencing
• Copy Number Variant Analysis
• Germline Variant Filtering
• Variant Recalibration