Single day long read whole genome sequencing for Acute Myeloid Leukemia

Abstract Introduction: Acute myeloid leukemia (AML) is considered a medical emergency and its management has evolved dramatically over the last decade. Molecular and cytogenetic testing for complete genomic profiling are important factors in treatment decisions in patients with AML given the availability of targeted therapy. However, current testing methods for these therapies have varying turnaround times ranging from 3-14 days. Shorter reporting time while not compromising genomic content analyzed, is imperative to appropriate diagnosis and early therapy selection. Herein we present a comprehensive genomic profiling test that provides results in ~1 day from sample collection to reporting using whole genome sequencing. The clinical development of this test will prevent delays in treatment, provide a total assessment of genomic anomalies and has the potential to improve patient outcomes. Methods: 20 residual blood/bone marrow samples from patients diagnosed with AML (>16 percent blasts) were obtained which were known to have 17 insertions/deletions, 60 single nucleotide variations (SNV) with a allele frequencies ranging from 5 percent to 84 percent across 92 genes which are clinically relevant in AML. These included 5 clinically relevant translocations/inversions. DNA was extracted using the EZ1(Qiagen, Germany) and sheared to 10-20kb using g-TUBEs (Covaris, USA). DNA was quantified using Qubit 2.0 (Thermofisher, USA) with an average DNA concentration of 98ng/ul (17-178ng/ul). Whole genome libraries were prepared using the ligation sequencing kit V14 from Oxford Nanopore Technologies (ONT) (Oxford, UK). Libraries were sequenced on the PromethION 2 or PromethION 24 across 3 simultaneous flow cells with a data generation target of 160GB. Average sequencing time was 17 hours (15-20 hours) (50-60x coverage). Basecalling was performed using Dorado and alignment was performed using the minimap2 long read aligner. Variant analysis was performed using optimized custom models for SNV and structural variation calling using AUGMET, a comprehensive bioinformatics platform for genomic data. Results: Average time from sample receipt to data availability determined through addition of individual steps in a retroactive manner showed a total processing time of 24-36 hours including 3 hours of hands on time. 62/62 variants detected by clinical exome sequencing with an allele frequency > 15 percent were identified accurately in this cohort. 9/15 variants which were clinically reported below 15% VAF were also detected. No therapeutically actionable variants were missed. Moreover, all previously reported clinically relevant structural rearrangements including PML::RARA (t(15;17)(q24;q21)), inversion 16 (inv(16)), inversion 3 (inv(3)), MLLT3::KMT2A(t(9;11)(p22;q23) and KMT2A::AFDN (t(6;11)(q27;q23)) were detected. Conclusion: This data demonstrates that rapid long read whole genome sequencing can be used as a potential diagnostic test that provides results in ~1 day. Whole genome sequencing is sensitive enough to detect structural rearrangements, single nucleotide variants and insertions/deletions in samples with a blast count >20% that were detected at a variant allele fraction of >15%. Given the criticality of determining actionable variants at the time of diagnosis, which are generally dominant clones, this methodology will provide a complete assessment of a clinical whole genome. Integration of methylation calling, natively available in ONT sequencing without additional cost, promises a paradigm shift in routine analysis of AML and therapy determination given the extensive work performed on methylation classifiers for therapy. In addition to this, given the ease of library preparation and low cost of investment, this is a viable approach for large scale decentralized testing.