SCOTCH – isoform-level characterization of gene expression through long-read single-cell RNA sequencing

rna-seq

Single-cell RNA sequencing helps researchers understand which genes are active in individual cells. However, many traditional methods rely on short sequencing reads, which can make it difficult to tell which full-length RNA isoforms are present. Isoforms are different RNA versions produced from the same gene, and they can have very different biological functions.

Researchers at the University of Pennsylvania and the Raymond G. Perelman Center for Cellular and Molecular Therapeutics at Children’s Hospital of Philadelphia, developed a new computational pipeline called SCOTCH. The tool is designed to analyze long-read single-cell RNA sequencing data and identify full-length transcript isoforms more accurately.

Long-read sequencing technologies, such as Nanopore and PacBio, can read much longer RNA molecules than short-read platforms. This makes them especially useful for studying isoforms, because they can capture more complete transcript structures. SCOTCH was designed to work across different sequencing technologies and single-cell protocols, including 10X Genomics and Parse Biosciences.

One of the main challenges in isoform analysis is that sequencing reads can map ambiguously to more than one transcript. SCOTCH addresses this by modeling isoforms as combinations of smaller, non-overlapping transcript segments called sub-exons. This helps the pipeline assign reads more reliably and distinguish between known and novel isoforms.

Workflow of the SCOTCH pipeline for long-read scRNA-seq analysis

(a) For either the 10X Genomics or Parse Biosciences single-cell library construction method, the cell barcodes and UMIs are extracted using vendor-supplied protocols. SCOTCH accounts for potential read truncations depending on primers used. (b) SCOTCH dissects known isoform annotations into non-overlapping exons and integrates read coverage information from BAM files to refine these annotations by identifying novel candidate exons or splitting existing exons into sub-exons. SCOTCH then maps reads to each sub-exon, applying a threshold where reads are considered mapped if it covers more than 60% of the exon’s length, and considered unmapped if it covers less than 10%. Using these read-exon mapping calls, reads are then aligned to known isoforms through a compatible matrix. (c) Reads that cannot be mapped to any known isoforms are iteratively clustered to generate candidate novel isoform annotations at the pseudo-bulk level based on read-exon matching profiles followed by read-isoform remapping. (d) lr-scRNA-seq data generated using 10X Genomics or parse libraries can be preprocessed by vendor-supplied tools, such as the wf-single-cell or Parse pipelines for barcode identification and then processed by SCOTCH to generate count matrices at both the gene and transcript levels. The gene-level count matrix facilitates conventional gene expression analysis, while the isoform-level count matrix can be used for transcript usage analysis through SCOTCH’s statistical pipeline.

The method also uses dynamic thresholding and iterative clustering to improve isoform reconstruction. In simpler terms, SCOTCH adjusts its decision-making based on the data and repeatedly refines transcript structures. This allows it to recover novel isoforms that other methods may miss while reducing false-positive results.

The researchers tested SCOTCH using simulations and real datasets, including human blood and cerebral organoid data. Across these analyses, SCOTCH improved quantification of known isoforms and helped identify experimentally supported novel transcripts.

This matters because different cell types may use different isoforms from the same gene. By resolving these isoform-level differences, SCOTCH can provide a more detailed view of cell identity, gene regulation, and disease-related biology than gene-level analysis alone.

As long-read single-cell RNA sequencing becomes more widely used, tools like SCOTCH will be important for turning complex sequencing data into meaningful biological insight. Better isoform analysis may help researchers understand how cells specialize, how tissues develop, and how transcript changes contribute to disease.

Availability – The SCOTCH software developed in this study is publicly available at: https://github.com/WGLab/SCOTCH

Xu Z, Qu H, Chan J, Mu S, Kao C, Hakonarson H, Wang K. (2026) SCOTCH: isoform-level characterization of gene expression through long-read single-cell RNA sequencing. Nature Communications [Epub ahead of print]. [article]