Clonal Hematopoiesis Filter

The gold-standard computational strategy that requires sequencing a paired buffy coat or peripheral blood mononuclear cell (PBMC) sample alongside the plasma cfDNA. By genotyping the patient's own white blood cells, the algorithm directly identifies and subtracts somatic variants originating from the hematopoietic lineage. This method definitively distinguishes a JAK2 V617F mutation arising from CH from an identical mutation shed by a solid tumor.

An emerging machine learning strategy that infers the cell of origin directly from the physical properties of cfDNA fragments without a matched blood control. CH-derived DNA retains the nucleosome protection patterns of myeloid cells, while tumor-derived DNA exhibits distinct fragmentation profiles. Algorithms analyze:

• Fragment length distributions: Hematopoietic cfDNA is typically longer than tumor cfDNA.
• End motif frequencies: The specific nucleotide sequences at fragment ends differ by tissue of origin.
• Nucleosome positioning: The spacing of protected DNA wrapped around histones acts as a cell-type signature.

A simple heuristic that leverages the biological principle that CH clones often contribute a high fraction of total leukocytes, while early-stage tumor ctDNA is present at very low concentrations. By applying a high VAF cutoff (e.g., > 10% in the absence of a matched tumor), variants likely representing clonal expansions are computationally suppressed. This method is computationally cheap but risks filtering true tumor mutations in patients with high tumor burden or masking CH variants that have not yet expanded to a high VAF.

A monitoring strategy that distinguishes CH from tumor recurrence by observing variant kinetics over multiple time points. CH clones typically exhibit stable or slowly increasing VAFs over months to years, reflecting the steady expansion of a hematopoietic stem cell. In contrast, tumor-derived ctDNA variants show rapid, exponential increases correlating with disease progression or sharp declines following successful therapy. This temporal analysis is critical for minimal residual disease (MRD) assays to prevent a stable CH clone from being mistaken for treatment-refractory cancer.

The computational subtraction of inherited polymorphisms from a somatic variant callset by comparison against a matched normal sample or population database. Unlike clonal hematopoiesis filtering, germline filtering removes variants present in every cell from conception, not just blood-specific clones.

• Uses matched normal DNA from buccal swabs or buffy coat
• Population databases like gnomAD provide allele frequency baselines
• Essential preprocessing step before somatic variant calling

A curated collection of sequencing data from healthy individuals used to model and suppress systematic technical artifacts and recurrent background noise in somatic variant callers. PoNs capture site-specific sequencing errors and platform-specific biases that mimic low-frequency variants.

• Built from age-matched, cancer-free donors
• Models position-specific error rates across the capture space
• Complements CH filtering by removing technical rather than biological noise

The percentage of sequencing reads at a specific genomic locus that contain a variant allele. In liquid biopsy, low VAF variants (typically <2%) are the primary challenge—they may represent true ctDNA from a tumor, clonal hematopoiesis, or sequencing artifacts.

• CH variants often present at VAFs of 0.5-10%
• Tumor-derived ctDNA VAFs can be orders of magnitude lower
• VAF alone cannot distinguish CH from tumor origin

The gold-standard biological approach to CH filtering: deep sequencing of peripheral blood mononuclear cells alongside plasma cfDNA. Variants present in both compartments are classified as CH-derived and excluded from the ctDNA callset.

• Requires paired buffy coat or PBMC sample
• Enables definitive CH attribution through shared variant detection
• Adds cost and complexity but dramatically reduces false positives

A computational method that analyzes the trinucleotide context of detected variants to infer their mutagenic origin. CH-associated mutations often carry signatures of aging (SBS1/SBS5) or specific exposures, while tumor variants may display signatures of APOBEC activity (SBS2/SBS13) or DNA repair deficiency.

• COSMIC mutational signatures catalog provides reference patterns
• Signature analysis can flag likely CH variants without matched blood
• Emerging as a probabilistic alternative to paired sequencing

A curated list of genes recurrently mutated in clonal hematopoiesis of indeterminate potential (CHIP), including DNMT3A, TET2, ASXL1, PPM1D, TP53, JAK2, and SF3B1. Variants in these genes, especially at low VAFs, are automatically flagged for CH filtering review.

• DNMT3A and TET2 account for the majority of CH mutations
• PPM1D mutations are enriched after chemotherapy exposure
• Gene-level filtering reduces but does not eliminate CH false positives