INTRODUCTION: Assessment of MRD in AML is emerging as an important tool for refining risk classification and improving patient management and outcomes. MRD also shows promise as a surrogate endpoint to accelerate therapeutic development and may facilitate more robust post-transplant surveillance. Highly sensitive digital PCR (dPCR) methods such as BEAMing - which are 50-100X more sensitive than "pan-heme" NGS assays - have generated compelling data for therapeutic response monitoring and clearance of molecular MRD in clinical trials: for example, for both relapsed/refractory as well as newly diagnosed patients, declining levels and eventual clearance of IDH1 mutations correlated with more robust hematologic response (DiNardo CD et al. N Engl J Med. 2018, 378(25):2386-98; Daigle S et al. Presented at ASH 2019). Nevertheless, the complex clonal dynamics of AML demand broader coverage across more genomic regions than are currently able to be interrogated using dPCR. At the same time, NGS pan-heme panels that offer broad genomic coverage lack sufficient sensitivity to detect MRD, as their limits of detection (LoD) are between 1-5% mutant allele frequency (MAF). To overcome these limitations, we developed a highly sensitive assay based on the Safe-SeqS technology (SafeSEQ) to detect de novo mutations across genomic regions with established and emerging clinical validity for AML MRD. This highly sensitive error-corrected NGS-based workflow detects molecular MRD with an LoD comparable to dPCR assays such as BEAMing. METHODS: A multiplexed enrichment panel targeting 68 highly relevant genomic regions was developed for ultra-sensitive mutation detection across 57 exons spanning 20 genes: BCOR, BRAF, CEBPA, FLT3, GATA2, IDH1, IDH2, JAK2, KIT, KRAS, NPM1, NRAS, PRPF8, PTPN11, SETBP1, SF3B1, SRSF2, TP53, U2AF1, and ZRSR2. Initial analytical performance demonstration for the SafeSEQ AML MRD test examined limit of blank (LoB) as well as LoD at the 95% confidence level (LoD95). LoB was measured using wildtype (wt) gDNA prepared from two NIST cell lines run in replicate; LoD was determined through replicate testing of contrived materials containing 30 characterized mutations across different mutant molecule (MM)/ MAF tiers ranging from 20 MM down to <1 MM. Low and high DNA input levels were also tested, from 1,000 genomic equivalents (GE) up to a subset of tiers run at 20,000 GE (3.3-66 ng) to demonstrate performance irrespective of wt DNA amount. Direct evaluation of analytical and clinical performance between the SafeSEQ AML MRD panel and a subset of mutations detected using BEAMing is currently underway. RESULTS: A rank-based approach was used to evaluate the distribution of background signals observed in the wt cell line DNA across all interrogated base positions. Consistent with previous observations of LoB across different SafeSEQ panels, a very low level of background noise was observed, with the vast majority of bases exhibiting signal <1 MM, indicating robust analytical specificity. Importantly, there were no positions affected by elevated noise for which clinically relevant mutations have been characterized for AML. For the variant-positive contrived materials run in replicate across different mutation levels, hit rates were consistent with those observed in other SafeSEQ assays and yielded 95% detection at 5 MM; this corresponds to 0.025% and 0.5% MAF for 20,000 and 1,000 GE inputs, respectively. Hit rates for the different MM tiers were also consistent between low and high DNA input levels, indicating robust performance for mutation detection across a wide range of background wt DNA. CONCLUSIONS: The SafeSEQ AML MRD assay demonstrates ultra-sensitive detection of low frequency mutations, with an LoD95 of 5 MM (0.025% MAF for 20,000 GE DNA input), while specificity remains very high. Similar to other SafeSEQ platform configurations, this performance is comparable with data obtained using the BEAMing dPCR technology; it is also orders of magnitude more sensitive than broad pan-heme NGS tests. With this combination of ultra-high analytical sensitivity paired with clinically-informed genomic coverage, we anticipate the SafeSEQ AML MRD test will play a fundamental role in accelerating therapeutic development. It should also provide clinicians with more reliable MRD information at important decision points, such as pre-transplant, or when considering novel maintenance therapies. Disclosures Sloane: Sysmex Inostics, Inc: Current Employment.