Variant Allele Frequency (VAF)

VAF is calculated as the ratio of variant-supporting reads to the total read depth at a specific genomic position.

• Formula: VAF = (Variant Reads) / (Reference Reads + Variant Reads)
• Range: 0.0 (absent) to 1.0 (homozygous), though somatic variants in bulk tissue typically fall below 0.5 due to stromal admixture.
• Example: A mutation observed in 150 out of 1,000 total reads yields a VAF of 0.15 (15%).

VAF is directly modulated by sample purity and the zygosity of the mutation. In a 100% pure tumor sample, a heterozygous somatic mutation has an expected VAF of 50%, while a homozygous mutation approaches 100%.

• Purity Correction: Adjusted VAF = Observed VAF / Tumor Purity
• Subclonal Mutations: Variants present in only a fraction of tumor cells exhibit lower VAFs, forming the basis for reconstructing subclonal architecture.
• Copy Number Impact: Gains or losses at the variant locus distort the expected VAF, requiring integration with Copy Number Alteration (CNA) calls.

In plasma-derived Circulating Tumor DNA (ctDNA) analysis, VAFs are often extremely low, reflecting the minute fraction of tumor-derived DNA amidst a vast background of wild-type Cell-Free DNA (cfDNA).

• Detection Thresholds: Modern assays using Unique Molecular Identifiers (UMIs) and Duplex Sequencing can reliably detect variants at VAFs below 0.1%.
• Monitoring Applications: Serial VAF measurements track treatment response, with a decreasing VAF indicating tumor regression and a rising VAF signaling molecular relapse or acquired resistance.
• Technical Challenges: Low VAFs require rigorous Base Quality Recalibration (BQR) and Germline Filtering to distinguish true signal from sequencing noise.

A critical pitfall in VAF interpretation is Clonal Hematopoiesis of Indeterminate Potential (CHIP). Age-related somatic mutations in hematopoietic stem cells are shed into the plasma alongside tumor-derived ctDNA.

• Misattribution Risk: A CHIP variant can be mistaken for a tumor-derived mutation, leading to incorrect therapy selection.
• Mitigation Strategy: A Clonal Hematopoiesis Filter requires parallel sequencing of matched peripheral blood mononuclear cells (PBMCs) to computationally subtract hematopoietic variants from the plasma variant callset.
• VAF Clues: CHIP mutations often exhibit stable, low VAFs that do not correlate with tumor dynamics.

The distribution of VAFs across multiple somatic mutations within a single tumor sample encodes its subclonal architecture. Mutations shared by all cancer cells (clonal) cluster at higher VAFs, while mutations in subpopulations (subclonal) form lower-VAF clusters.

• Cancer Cell Fraction (CCF): VAF is converted to CCF by correcting for purity and local copy number, estimating the proportion of cells harboring the mutation.
• Phylogenetic Reconstruction: Clustering mutations by CCF allows inference of the tumor's evolutionary history and identification of truncal driver mutations present in all cells.
• Clinical Relevance: Targeting clonal truncal mutations minimizes the risk of resistance from pre-existing subclonal populations.

Accurate VAF estimation is susceptible to multiple technical biases that must be computationally corrected.

• GC Bias: Non-uniform coverage due to fragment GC content distorts read depth and VAF. GC Bias Correction normalizes coverage profiles.
• Mapping Quality: Reads mapping to multiple genomic locations (low MAPQ) can artificially inflate or deflate variant counts.
• Index Hopping: Sample barcode misassignment during sequencing creates cross-sample contamination, introducing spurious low-VAF variants.
• Library Complexity: Low-complexity libraries with excessive PCR duplicates reduce the effective depth and statistical power for VAF estimation, necessitating Molecular Barcode deduplication.

A random nucleotide barcode ligated to individual DNA molecules before PCR amplification. UMIs enable computational deduplication by grouping reads sharing the same barcode, collapsing them into a single consensus read. This process corrects polymerase errors and eliminates amplification bias, enabling accurate VAF calculation from the original template molecule count.

• Without UMIs: VAF is distorted by PCR duplication and sequencing errors
• With UMIs: VAF = (consensus mutant molecules) / (total consensus molecules)

The lowest VAF at which a variant can be reliably distinguished from background noise with predefined confidence (typically 95%). LoD is determined by sequencing depth, error rate, and input DNA quantity. A typical targeted panel with 5,000x depth may achieve an LoD of 0.1% VAF, while duplex sequencing can push detection below 0.01%.

• Formula: LoD ≈ (3 × √background_error_rate) / sequencing_depth
• Clinical threshold: Most MRD assays target LoD ≤ 0.01% VAF for post-treatment surveillance

A specialized algorithm that distinguishes true low-frequency somatic mutations from germline variants, sequencing errors, and mapping artifacts. These callers model per-base error rates, strand bias, and position-specific noise to assign a statistical probability that a non-reference allele at a given VAF is real.

• Key algorithms: MuTect2, VarScan2, Strelka2, LoFreq
• Critical output: Variant allele fraction with confidence intervals, not just a binary call
• Tumor-only mode: Requires Panel of Normals to filter systematic artifacts when no matched normal is available

The inferred population structure of genetically distinct tumor cell clusters reconstructed from the distribution of VAFs across multiple somatic mutations. Mutations present in all cancer cells (clonal) exhibit higher VAFs, while those in a subset (subclonal) appear at lower frequencies. This analysis reveals tumor evolution and heterogeneity.

• Clonal mutation: VAF ≈ (tumor purity × copy number) / (2 × (1 - purity) + purity × total copy number)
• Subclonal mutation: VAF is further reduced by the fraction of cells harboring the mutation
• Clinical implication: Subclonal driver mutations may emerge under therapeutic selective pressure

A computational or matched-control strategy to exclude somatic variants originating from age-related clonal expansions in blood cells (CHIP) rather than from a solid tumor. Without filtering, CHIP variants at low VAF can be misattributed as ctDNA, leading to false-positive liquid biopsy results.

• Filtering strategies: Matched buffy coat sequencing, population databases, or machine learning classifiers using fragmentomics features
• Common CHIP genes: DNMT3A, TET2, ASXL1, PPM1D
• Prevalence: Affects >10% of individuals over age 70, making this filter essential for accurate VAF interpretation