RAMBO: Resolving Amplicons in Mixed Samples for Accurate DNA Barcoding with Oxford Nanopore

DNA barcoding, the use of short genetic markers to identify and differentiate species, is a foundational tool for ecological and taxonomic research. The method has been scaled rapidly with next-generation sequencing technologies enabling the processing of thousands of specimens in parallel. Nanopore sequencing not only offers a flexible, low cost alternative to other platforms but produces full-length reads in real time and can be used in remote settings. However, its comparatively high error rate complicates downstream processing, particularly when PCR amplifies multiple templates from a single specimen, reflecting pseudogenes, paralogs, or contaminants. We present a novel pipeline for DNA barcoding that resolves mixed sequence signals from Nanopore reads using unsupervised clustering and staged consensus generation, without relying on curated reference databases, taxonomic priors, or error models. While existing methods to curate Nanopore sequence data assume a single dominant amplicon per sample or require deep sequence divergence among amplicons, our pipeline can distinguish variants differing by as little as 0.15 percent. It combines column-weighted encodings, UMAP projection, and HDBSCAN clustering, followed by conservative consensus refinement. The pipeline was benchmarked and validated using datasets with known composition, including high-fidelity PacBio sequences. The results show that Nanopore barcoding, when paired with appropriate analysis, can recover biologically meaningful variation even in technically complex samples. The pipeline is particularly suited for specimens where divergent templates are co-amplified, including mitochondrial pseudogenes or multicopy nuclear regions like ITS. As such, it provides a generalizable framework for high-resolution Nanopore analysis of complex amplicon mixtures.