Probe-Free Multiplexed RPA Detection Via Single-Molecule Nanopore Sensing and Deep Learning Classification

Abstract

Recombinase Polymerase Amplification (RPA) is a rapid, sensitive, and isothermal method for nucleic acid amplification that has gained widespread use in diagnostic applications. While well-suited for single-target assays, extending RPA to multiplex detection remains technically challenging despite the multiplexed RPA being essential for tasks such as differential diagnosis and inclusion of internal controls. Existing multiplex RPA strategies rely on proprietary probes (e.g., Exo, Fpg, Nfo), which require complex design, are susceptible to cross-reactivity, and often depend on sophisticated optical instrumentation, limiting their scalability. To address these limitations, here, we developed a probe-free multiplex RPA assay that distinguishes targets by amplicon length, using solid-state nanopore detection of single molecules and deep neural network (DNN)-based classification. Using the Monkeypox (Mpox) as a model, we designed and validated a multiplex RPA assay targeting both the Mpox gene and the human RNase P gene as an internal control. We systematically evaluated the nanopore size requirement for detecting DNA amplicons ranging from 50 to 500 base pairs and found that [~]7 nm pores provided optimal performance for amplicons between 75 and 500 bp, balancing high event rates without pore clogging. Using single-molecule translocation events as input, we trained and optimized a DNN to classify amplicons by target. The model achieved 94.2% accuracy at the single-molecule event level, which translated to 100% accuracy at the population-level target call. While our current system demonstrates duplex detection, the strategy is inherently scalable to higher multiplex levels. These findings establish a new framework for multiplex RPA that eliminates the need for complex probe design and optical detection, paving the way for robust, scalable, and accessible molecular diagnostics.