Multi-Nucleotide Variants Enhance Circulating Tumor DNA Detection in B-cell Lymphoma
Minimal residual disease (MRD) detection using circulating cell free DNA (cfDNA) of B-cell lymphoma patients is hampered by the low fraction of circulating tumor DNA (ctDNA) and NGS error rates. B-cell lymphomas often harbour clustered AID-induced single nucleotide variants (SNVs) in hypermutated super-enhancer regions. Together, these clustered SNVs form multi-nucleotide variants (MNVs) which offer an opportunity to improve limit of detection.
We developed MNVista, a bioinformatics tool that detects MNVs starting from SNVs called by an in-house developed pipeline and the corresponding aligned sequencing reads (indexed BAM files). The algorithm constructs genomic windows, pairs proximal SNVs, and statistically evaluates their co-occurrence using a Bayesian framework supported by Phi-coefficient and Jaccard index metrics. MNVs consisting of >2 SNVs are subsequently assembled by iteratively merging mutually compatible pairs.
Using in-silico spike-ins of synthetic MNVs at VAFs 0.1–10% into five DLBCL cfDNA BAMs (~18,000× median coverage) and realistic somatic background noise, MNVista achieved F1 scores >0.97 and PR-AUCs >0.94 in semi-synthetic settings (Youden-derived Bayesian thresholds >0.99), and reliably detected spiked MNVs down to <0.1% VAF (with theoretical sensitivity limits reaching <1 ppm for larger MNVs). In longitudinal patient samples, baseline MNVs aligned closely with clinical outcome: patients who harbored somatic MNVs at interim and end-of-treatment (EOT) timepoints at VAFs of ~0.01–0.1% relapsed, despite achieving PET/CT-defined complete remission (CR). Conversely, patients lacking detectable somatic MNVs at EOT remained in CR. These results indicate that MNV-based ctDNA calling increased sensitivity of MRD detection and outperformed imaging in identifying patients at risk to develop a relapse.
MNVista leverages the low joint error probability of co-occurring in-cis variants to substantially enhance ctDNA sensitivity and is compatible with standard high-depth, UMI-aware pipelines. MNVista can be readily integrated for MRD monitoring and prospective clinical studies in (germinal center derived) B-cell lymphoma with localized hypermutated regions.
