A systematic benchmark of bioinformatics methods for single-cell and spatial RNA-seq nanopore long reads data

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Alternative splicing significantly contributes to transcriptomic complexity.
• Resolving alternative splicing at the single-cell level is challenging with short-read sequencing.
• Long-read sequencing offers full-length transcript information crucial for accurate isoform detection.

Purpose of the Study:

• To benchmark computational tools for single-cell and spatial long-read transcriptomics.
• To compare the effectiveness of different analytical approaches for isoform detection and gene expression profiling.
• To provide a reusable resource for evaluating bioinformatics methods in long-read transcriptomics.

Main Methods:

• Generation of paired short-read and Nanopore long-read single-cell datasets.
• Evaluation of ten state-of-the-art computational methods.
• Benchmarking across four analytical dimensions: barcode/UMI detection, demultiplexing, gene expression profiling, and isoform analysis.
• Assessment using real and simulated datasets with varying protocols, sequencing depths, and chemistries.

Main Results:

• Identified method-specific trade-offs in accuracy, robustness, and scalability.
• Highlighted the critical impact of sequencing quality and unique molecular identifier (UMI) correction strategies.
• Demonstrated the superiority of long-read sequencing for full-length isoform detection compared to short-read methods.

Conclusions:

• This benchmark provides a practical resource for optimizing isoform analysis and gene expression profiling in single-cell and spatial transcriptomics.
• The findings guide the selection of appropriate bioinformatics tools for long-read transcriptomic data.
• The reusable benchmarking workflow facilitates ongoing development and comparison of new computational methods.