MIRAGE: A ctDNA methylation-driven computational algorithm designed for sensitive detection of minimal residual disease

Background: Detection of minimal residual disease (MRD) following surgical resection or a treatment regimen remains a significant clinical challenge, primarily due to the typically low concentrations of circulating tumor DNA (ctDNA) in the bloodstream post-treatment. In situations where circulating tumor cells (CTCs) are undetectable or ctDNA levels fall below the thresholds of conventional detection methods, DNA methylation analysis has emerged as a promising alternative. Utilizing computational algorithms further improve the detection of tumor signal in cancers with as low as 0.001% circulating tumor fraction. Methods: MIRAGE (Minimal Residual Assessment using Genome-wide Epigenomics) algorithm was validated using 156 samples, among which 16.7% (n=26) were control reference samples. The algorithm assessed DNA methylation patterns across genome-wide differentially methylated regions (DMRs). The algorithm quantifies methylation by evaluating the number of methylated and unmethylated cytosines at each CpG site and integrating these values into a composite methylation score, reflecting overall hyper- or hypomethylation. The resulting methylation score is further normalized against a predefined cut-off established from a reference cohort of non-cancerous healthy individuals to enable standardized interpretation. Results: MIRAGE achieved a specificity of 96.8% from 127 clinical samples (64 tumor, 63 non-cancerous healthy controls) tested. Among the clinical tumor samples, 64% (n=41/64) were detected as ctDNA-positive which showed an overall sensitivity of 64.1%. Conclusion: The study demonstrates the specificity of MIRAGE algorithm for ctDNA classification-based minimal residual detection with high specificity. Further validation on larger cohort is still warranted to show more clinical insights.