Massively multiplex blocker displacement amplification (BDA) for the detection of measurable residual disease in acute myeloid leukemia.

Abstract

e19045 Background: Next-generation sequencing (NGS) technologies, used to accomplish high throughput characterization of DNA, have an inherent error rate of between 0.1-1%. However, many important DNA sequences possess variant allele frequencies (VAFs) below this rate, potentially precluding their detection and identification by NGS. The previously published blocker displacement amplification (BDA) research technology has enabled selective amplification of sequences with variants, facilitating detection by NGS of these sequences with low VAFs at low read depths. Methods: Here we present the massively multiplex BDA library preparation method for the NGS-based detection of low VAF sequences. By optimizing our previously published Simulated Annealing Design using Dimer Likelihood Estimation (SADDLE) algorithm, we designed a 4 tube assay, comprising up to ~2400 amplicons in each tube, and ~8300 total amplicons. Tube 1 was designed with the highest likelihood of on-target amplification and the lowest likelihood of dimerization or non-specific amplification. While Tube 1 targeted ~1000 regions covering the most-frequently observed mutations in current public databases of patients with acute myeloid leukemia (AML) across more than 200 genes, Tubes 2-4 were designed for enrichment of the full coding sequence of 63 genes. To highlight the performances of SADDLE, coverage uniformity, primer dimerization, and non-specific amplification of tube 1 were compared to those of a naively designed tube. The panel was further tested with contrived and reference samples to demonstrate its analytical performance. Results: The SADDLE algorithm design decreased the first pass amplicon dropout rate more than 10-fold, from 30% to ~2%, and further improved the overall on-target rate compared to the naively designed panel. Through SADDLE, we were able to achieve an 80% median coverage of 63 genes. The resulting panel has a bulk limit of detection (LoD) of below 0.1% VAF. Conclusions: Massively multiplex BDA enabled the NGS characterization of DNA targets with VAFs suitable for detection of AML measurable residual disease (MRD). This technology represents a 100-fold increase in current plexity for the detection of low VAF alleles over our previously published multiplex BDA technology, demonstrating the ability to perform high throughput sequencing of rare mutations.