Oxford Nanopore User Group Meeting, Chicago

Long-read single-cell RNA sequencing enables direct characterization of full-length transcript isoforms that remain unresolved by conventional short-read approaches. In this talk, I will present our development of Oxford Nanopore Technologies–based experimental and computational methods for high-throughput, isoform-resolved single-cell transcriptomics. Using human tissues, we constructed cell type-specific isoform atlases and identified coordinated changes in isoform usage associated with disease progression. These studies reveal how alternative transcript structures contribute to cellular identity and disease-associated state remodeling beyond changes in gene-expression levels alone. Together, our findings establish long-read single-cell sequencing as a powerful platform for discovering transcript diversity and its functional relevance in human disease.

In this clinical research study we performed single-cell long-read sequencing on immune cells from cerebrospinal fluid (CSF) and blood to assess isoform diversity in healthy aging individuals and those diagnosed with mild cognitive impairment (MCI) or Alzheimer’s disease (AD). cDNA from single-cell experiments was subjected to long-read sequencing using Oxford Nanopore Technologies. The dataset included immune cells from CSF (45 controls, 13 MCI/AD) and blood (22 controls, 28 AD). Computational analysis incorporated scNanoGPS for extracting cell barcodes and mapping reads to the genome, IsoQuant for isoform modeling, SQANTI3 for filtering artifacts, Hypatia for differential transcript usage (DTU) analysis, and ScIsoX for assessing isoform diversity. We identified 97,920 unique transcripts in CSF and 139,351 in blood. Approximately half of expressed genes in both compartments were expressed through multiple isoforms, and 20% of detected isoforms were previously unannotated in the human genome. In the CSF, several AD risk genes - including APOC1 and MS4A6A - expressed multiple isoforms within single cells, suggesting complex isoform-level regulation in immune populations. DTU analysis revealed significant splicing shifts in CSF-derived monocytes of individuals with cognitive impairment. We conclude that human immune cells demonstrate extensive isoform diversity, including previously unannotated isoforms, particularly in genes implicated in AD. Disease-associated splicing changes suggest that isoform-level regulation may contribute to immune dysfunction in AD.

Nanopore sequencing provides long-range genomic and transcriptomic information, but technical chimeras can create junctions that closely resemble genuine structural variants or transcript rearrangements. We have encountered this problem in several nanopore DNA and RNA sequencing workflows, where similar-looking read structures can arise through very different mechanisms.

In direct RNA sequencing, internal adapters can join otherwise unrelated RNA fragments and are difficult to recognize after basecalling. DeepChopper uses a genomic language model to identify these adapter sequences at nucleotide resolution and resolve the resulting chimeric reads. In direct-cDNA sequencing, the artifact landscape is broader because reverse transcription can generate foldback inversions, internal adapter chimeras, and homopolymer-mediated template switching. DirectClean was developed to detect these events while rescuing usable sequence for downstream transcriptome analysis. A related challenge occurs in whole-genome-amplified long-read DNA data, where amplification can generate chimeric reads that mimic structural variants. ChimeraLM uses matched WGA and bulk sequencing to distinguish these artifacts from genuine genomic events, while OctopuSV is used to harmonize and compare structural variant calls for downstream evaluation.

Across these projects, the common goal is to remove technical junctions without losing the long-range information that makes nanopore sequencing valuable for genomic and transcriptomic analysis.