Tonight I watched Gabriel Griffin, Assistant Professor, Dana-Faber Cancer Institute present an Oxford Nanopore Technology session titled “Rapid epigenomic classification of acute leukemia with long-read sequencing.” Griffin spoke about leukemia diagnosis requiring multiple steps and time: morphology, flow cytometry, immuno histochemistry stains, targeted sequencing, and cytogenetics. This can take up to fourteen days. Griffin and team hypothesized that epigenomics could be used for rapid leukemia classification. They explained that there is data on epitypes and roles of methylation in cancer. Nanopore long-read sequencing captures important features and methylation signals. The research team collaborated with another lab to produce a DNA methylation-based reference of acute leukemia using methylation array data from 11 studies and over 2,500 samples. The aggregated data was clustered and separates out leukemia types and molecular categories. The team trained a machine learning classifier for sparse methylation data. The model was trained by masking parts of datasets. The model was then validated in a retrospective cohort at Dana Faber Cancer Institute. With clinical colleagues, they then validated the MARLIN model for rapid/real-time diagnosis. They performed 20 minutes of DNA extraction and 20 minutes of library prep. With forty or fifty minutes of sequencing, the classifier reached the confidence threshold. A p53 mutation was found and confirmed with additional tests. Griffin shared additional case studies and spoke about future studies to improve sensitivity, specificity, and detection limits.
