Bioinformatics Frameworks for Single-Cell Long-Read Sequencing

Peer-reviewed review (Briefings in Bioinformatics, Dec 2025) on the convergence of single-cell resolution with third-generation (long-read) sequencing — termed SCLR-seq. The central claim: long reads now cover entire transcripts at single-cell resolution, so analysis can move beyond differential gene expression to differential isoform expression — the actual functional units of the transcriptome.

Accuracy (quoted figures)

• PacBio HiFi: read accuracies of 99.9% (Sequel II) and 99.95% (Revio), via circular consensus sequencing (CCS).
• Oxford Nanopore (ONT): reported accuracies up to ~99%, driven by newer flow-cell chemistry (R10.4) and improved basecalling — closing the historical accuracy gap with PacBio and short-read Illumina.

Throughput (quoted figures)

• Revio: 80–100 million HiFi reads per run.
• Vega: 50–60 million reads per run.
• MAS-Iso-Seq: ~40 million HiFi reads per run on the Sequel IIe.
• MAS-seq / Kinnex concatenate cDNA molecules from single-cell platforms (e.g. 10x Genomics) into long fragments compatible with PacBio long-read instruments, then bioinformatically de-concatenate — giving full-length isoform information per single cell where short reads give only gene-level counts.

Why Full-Length Isoform Detection Matters

• Short-read scRNA-seq is limited by read length and 3′/5′ capture bias, preventing accurate full-length isoform reconstruction.
• SCLR-seq gives an annotation-independent view revealing the full extent of isoform diversity, co-occurring splicing events within a single isoform, coordinated alternative-promoter/splicing patterns, novel isoforms, and fusion transcripts — all invisible to fragmented short-read data.

Clinical / Disease Applications

• Cancer: CD44 splice variants in pancreatic and breast cancer metastasis; neoepitope/MHC-I discovery from alternative-splicing events in breast and ovarian cancers (a route to splicing-derived immunotherapy targets).
• Neurological: cell-type-specific splicing patterns in autism.

Bioinformatics Tooling (the field is maturing)

• End-to-end pipelines: FLAMES, SiCeLoRe, nanoseq, scywalker.
• Barcode/UMI extraction: BLAZE, Flexiplex, scTagger.
• Isoform quantification: Isosceles, SCOTCH, lr-kallisto, Bambu-clump.
• Downstream: SQANTI3 (classification), scisorseqr (differential splicing), ScisorWiz (visualization).

Why It Matters for the KB

Long-read sequencing is the read-out layer beneath gene editing: validating edits, mapping variant→isoform→function, and building the single-cell atlases the precision-longevity thesis depends on. With ONT at ~99% and PacBio HiFi at ~99.95%, the "accuracy tax" that kept long reads out of the clinic is largely paid down — single-cell + long-read is now a production tool, not a frontier promise.

Source

• Bhatia S., Field M.A., Hebbard L., Schmitz U. "Bioinformatics frameworks for single-cell long-read sequencing: unlocking isoform-level resolution." Briefings in Bioinformatics, 26(6), bbaf655, Dec 11 2025. DOI 10.1093/bib/bbaf655.