Towards hyper-sensitive MRD: Assessment of Enspyre as a cost-effective ctDNA detection technique using very large variant panels.

Abstract

e15048 Background: Tumor-informed minimal residual disease (MRD) assays track patient-specific variants at significantly higher sensitivity than tumor-naïve approaches, reaching limits of detection below 10 parts per million (ppm). However, for indications such as NSCLC, sensitivity remains a challenge: even the most sensitive assays fail to achieve 100% negative predictive value for recurrence. Sensitivity scales with number of variants tracked, molecule recovery efficiency, and DNA input mass. NSCLC tumors typically harbor >30,000 mutations so larger panel size remains a path to improved sensitivity, but incurs significant increases in costs. Enspyre is a novel tumor-informed MRD approach which reduces sequencing requirements by as much as 100-fold, facilitating use of significantly larger panels at lower costs. We present an assessment of Enspyre’s performance with panels of up to 30,000 variants. Methods: Fragmented cell line DNA was diluted at different ratios into variant-free samples. Panels of up to 30,000 variants comprising both SNVs and InDels present in the cell line were identified, and probes designed to target each variant. Variant-free DNA was used to generate data on background noise levels using independent probe pools, which was then used to inform error priors in ctDNA analysis models. Sensitivity was estimated using molecule counts along with a Probit regression approach with different panel sizes randomly selected from the 30,000 variant panel. Additionally, in silico down-sampling of variants and reads respectively was used to assess the effect of variant number and/or sequencing depth on sensitivity. Specificity was assessed using fragmented variant-free DNA. Results: Sensitivity was found to scale pseudo-linearly with panel size, both in direct measurement and down-sampling analysis, supporting the use of larger panels to increase sensitivity. Read subsampling showed consistent performance at sequencing depths down to ~2% of that used by current approaches, enabling use of significantly larger variant panels while decreasing sequencing requirements and reducing costs. No false positives were observed from variant-free samples, implying 100% specificity of the assay. Using panel sizes of 10,000 variants, ctDNA was consistently detectable above background levels at 1ppm, while panel sizes of 20-30,000 variants enabled consistent detection at or below 0.5ppm. Conclusions: These results demonstrate the potential of Enspyre to enable significantly larger tumor-informed MRD panels while simultaneously lowering sequencing requirements. This opens doors to a next generation of tumor-informed MRD assays with simultaneously higher sensitivity and lower costs, helping to address a significant challenge in the field.