Transcriptomics

Long-read RNA-seq analysis: full-length isoform catalogs from HiFi or Nanopore reads to filtered GTF

Long-read RNA sequencing can resolve full-length isoform structures that short-read RNA-seq often cannot reconstruct from fragments [1]. Pepkio runs version-pinned analysis from FASTQ or BAM to SQANTI3-filtered GTF catalogs. Custom inputs and workflows are scoped at kickoff. For academic, biotech, and pharma clients using PacBio Kinnex, typical targets are 5-10 million HiFi reads per sample [2], with scripts, figures, and a Methods draft.

Key facts

Key facts about Long-Read RNA-seq
FactValue
Supported platforms / instrumentsPrimary: PacBio Revio / Sequel II (Kinnex Iso-Seq HiFi). Oxford Nanopore PromethION / MinION (direct RNA, amplification-free cDNA, PCR-cDNA). MGI CycloneSEQ-WT02 / WY01 (nanopore-class, when scoped at kickoff). Single-cell long-read (MAS-ISO-seq, 10x-compatible cDNA) on request
Input requirements≥300 ng total RNA, RIN >7, DNase-treated; ≥3 biological replicates recommended for differential expression; 5M HiFi reads/sample (high-abundance isoform discovery), 10M (moderate-to-rare isoforms or isoform-level DE) for PacBio Kinnex [2, 3]
Reference builds supportedHuman GRCh38 + GENCODE v44; mouse GRCm39 + GENCODE vM33; custom references on request
Primary tools (with versions)minimap2 2.31; IsoQuant 3.13.0; SQANTI3 6.0.1; isoseq3 (SMRT Link 13.1); NanoPlot 1.42.0; DESeq2 1.44.0; DRIMSeq 1.30.0; stageR 1.28.0
Typical turnaround time4-6 weeks (standard bulk cohort, reference-guided); 6-10 weeks (annotation-free, multi-condition isoform DE, or orthogonal short-read validation). Confirmed at kickoff
Deliverable formats.gtf/.gff3, .bam, count matrices (.tsv/.csv); PDF/SVG figures; HTML QC report; documented R/Python scripts; Methods draft
Key cited best-practice referenceAmarasinghe et al. [1], Genome Biology; Pardo-Palacios et al. [4], Nature Methods (SQANTI3); LRGASP Consortium [5], Nature Methods
Custom / bespoke analysisNon-standard inputs, outputs, and methods scoped at kickoff, for example pre-aligned BAM, client GTF, fusion prioritization, ORF prediction, short-read junction validation, or single-cell long-read extensions

What is long-read RNA sequencing?

Long-read RNA sequencing maps individual RNA molecules that can span complete exon chains, so isoforms can be reconstructed rather than inferred from ~150 bp short-read fragments [1, 6]. Pepkio aligns PacBio HiFi or ONT reads with minimap2, builds transcript models with IsoQuant, and classifies isoforms as FSM, ISM, NIC, or NNC with SQANTI3 [4, 7, 8]. SG-NEx profiled seven human cell lines across five RNA-seq protocols and reported that long-read sequencing more robustly identifies major isoforms than short-read cDNA sequencing [6]. Custom deliverables outside the standard workflow are scoped at kickoff. See the long-read RNA-seq glossary.

When should you use long-read RNA sequencing?

Long-read RNA-seq is a fit when the question needs isoform-resolved expression, such as alternative splicing in disease, annotation of uncharacterized loci, or fusion validation. The table contrasts lrRNA-seq with bulk short-read RNA-seq and short-read scRNA-seq.

Comparison of long-read RNA-seq, bulk RNA-seq, and single-cell RNA-seq
ApproachBest forLimitationsApproximate cost range
Long-read RNA-seq (PacBio / ONT)Full-length isoform discovery, novel transcripts, fusion validation, ORF and UTR annotationHigher per-sample sequencing cost than bulk short-read; artifact-rich catalogs need curation; lower sample throughputLibrary prep, sequencing, and bioinformatics vary by platform, Kinnex multiplexing, and depth
Bulk short-read RNA-seqCondition-level gene DE on well-annotated genomesOften cannot resolve overlapping isoforms from junction fragments aloneLower per-sample cost for modest cohorts
Single-cell RNA-seq (short-read)Cell-type heterogeneity and rare populations3′ capture bias; typically no native full-length isoform resolutionMid-range; per-cell cost exceeds bulk long-read
  • Aged human brain isoform diversity: Aguzzoli Heberle et al. [9] mapped medically relevant isoform diversity in aged frontal cortex missed by short-read annotation.
  • Cross-platform benchmarking: Chen et al. [6] profiled SG-NEx across seven cell lines and five protocols, showing long-read data more robustly identifies major isoforms than short-read cDNA.
  • Pipeline artifacts: Du et al. [10] showed isoform-caller choice strongly affects false-positive novel isoform rates, which supports SQANTI3 curation before inference.

How the analysis works — step by step

  1. 1. Validate inputs and sample metadata

    Pepkio confirms FASTQ or BAM integrity, platform, chemistry, and metadata. Read depth, RIN, replicates, and contrast design are recorded in sample_manifest.csv. Sub-threshold yield or truncated reads are flagged before alignment [1, 2].

    Tools and outputs

    Tools used: fastqc / fastp as needed; samtools quickcheck for BAM inputs

    Output: sample_manifest.csv with library IDs, platform, read counts, and QC flags

  2. 2. QC raw long reads

    Read-length distributions, pass-filter rates, and full-length fractions are computed per sample. ONT reports include N50 read length and median quality. PacBio HiFi reports include mean read length and CCS pass rates. Truncation that inflates isoform catalogs is flagged [1, 4].

    Tools and outputs

    Tools used: NanoPlot 1.42.0; pycoQC 2.5.2 (ONT); PacBio dataset reports (HiFi)

    Output: read_qc_summary.csv; read-length histograms; per-sample QC flags

  3. 3. Process PacBio subreads to HiFi reads (when applicable)

    When clients deliver PacBio subreads or CCS BAMs, Pepkio runs the Iso-Seq workflow: demultiplex, refine poly-A tails, and cluster to HiFi reads [2]. Barcode crosstalk and low CCS yield are documented before alignment.

    Tools and outputs

    Tools used: isoseq3 (SMRT Link 13.1)

    Output: Demultiplexed HiFi .fastq.gz per sample; isoseq_stats.csv

  4. 4. Align reads to the reference genome

    Reads are mapped in splice-aware mode: minimap2 splice:hq for PacBio HiFi and splice with k-mer size 14 for ONT, with annotated GENCODE splice junctions supplied as BED input [7, 8]. Mapping rates, chimeric fractions, and primary vs. secondary alignments are audited per sample.

    Tools and outputs

    Tools used: minimap2 2.31; samtools 1.21

    Output: Coordinate-sorted, indexed .bam per sample; alignment_summary.csv

  5. 5. Reconstruct and quantify transcript models

    IsoQuant runs in reference-guided mode with GENCODE, extending the reference with sample-specific isoforms [7]. Gene-, isoform-, exon-, and intron-level counts are generated. Saturation is compared to platform depth targets [2, 6].

    Tools and outputs

    Tools used: IsoQuant 3.13.0

    Output: extended_annotation.gtf; isoform_counts.tsv; gene_counts.tsv; saturation curves

  6. 6. Collapse redundant isoform models

    Long-read pipelines often emit highly redundant transcript models that differ by terminal exons or indels ([4]; ConesaLab SQANTI3 wiki). When redundancy exceeds project thresholds, Pepkio collapses near-identical models before SQANTI3 classification.

    Tools and outputs

    Tools used: TAMA collapse or cDNA_Cupcake collapse_isoforms_by_sam as appropriate

    Output: collapsed_annotation.gtf; collapse audit log

  7. 7. Classify and filter with SQANTI3

    SQANTI3 assigns structural categories (FSM, ISM, NIC, NNC, genic, intergenic, antisense) and quality metrics on TSS, TTS, and splice junctions [4]. Rules-based filtering is the default. ML filtering is documented when selected.

    Tools and outputs

    Tools used: SQANTI3 6.0.1

    Output: SQANTI3_classification.txt; SQANTI3_filter_report.html; filtered corrected.gtf

  8. 8. Re-quantify the filtered transcript catalog

    SQANTI3 expression estimates are used for QC only, not for differential testing (ConesaLab SQANTI3 wiki; [4]). Pepkio re-runs IsoQuant quantification against the SQANTI3-filtered GTF per sample to produce final count matrices.

    Tools and outputs

    Tools used: IsoQuant 3.13.0 (--reference with filtered GTF)

    Output: filtered_isoform_counts.tsv; filtered_gene_counts.tsv; TPM tables

  9. 9. Test differential expression

    Gene-level contrasts use DESeq2 with Benjamini-Hochberg FDR correction when ≥3 biological replicates per condition are available. Isoform-level testing with DRIMSeq and stageR is run only when replicate count and read depth support isoform-resolved power [5, 10, 11]. Batch is included in the design matrix when the same contrast spans multiple sequencing runs [12].

    Tools and outputs

    Tools used: DESeq2 1.44.0; DRIMSeq 1.30.0; stageR 1.28.0

    Output: deg_results.csv; differential_isoform_usage.csv; MA and volcano plots

  10. 10. Package deliverables

    Pepkio assembles figures, exports count tables, writes commented scripts, and drafts a Methods section citing software versions. Custom milestones are included when scoped at kickoff.

    Tools and outputs

    Tools used: R 4.4.x / Python 3.12 scripts; ggplot2 3.5.1

    Output: Final deliverable bundle; HTML QC report; README; Methods draft

What Pepkio delivers

Processed data files

  • SQANTI3-filtered .gtf/.gff3; coordinate-sorted .bam + .bai; filtered_isoform_counts.tsv and filtered_gene_counts.tsv; SQANTI3_classification.txt; QC summaries.

Figures (PDF/SVG)

  • Read-length histograms; mapping-rate bar charts; gene/isoform saturation curves; SQANTI3 structural-category plots; MA and volcano plots when DE is in scope.

Tables

  • Key columns in SQANTI3_classification.txt (isoform, structural_category, associated_gene, coding, predicted_NMD); deg_results.csv (gene_id, log2FoldChange, padj); differential_isoform_usage.csv (isoform_id, gene_id, padj).

Code

  • Commented R and Python scripts per stage
  • Environment lock files (sessionInfo(), conda, or pip)
  • Delivery via private Git repository or agreed file transfer

Documentation

  • HTML/PDF QC report; README with reproduction instructions; Methods draft citing software versions and reference builds; post-delivery reviewer support for method clarification and minor revisions within agreed scope (typically ≤20% of deliverables).

Technical decisions we make — and why

Isoform reconstruction: IsoQuant (default) vs. FLAIR or Bambu
IsoQuant showed competitive F1-scores in LRGASP reference-guided benchmarks on annotated genomes [5, 7]. FLAIR or Bambu on request when clients require a specific caller or comparison.
SQANTI3 filtering: rules-based (default) vs. ML filter
Rules-based filtering is interpretable and reproducible. The ML filter is used when clients want more aggressive artifact removal [4].
Catalog build: pooled replicates vs. per-sample only
For discovery, Pepkio pools long-read samples to build a single experiment-level catalog, then re-quantifies per sample, matching the SQANTI3 recommended workflow (ConesaLab SQANTI3 wiki).
Reference-guided vs. annotation-free
Reference-guided with GENCODE is the default for human and mouse. Annotation-free discovery for non-model organisms is scoped separately, with orthogonal validation recommended [5].
Differential testing: gene-level DESeq2 vs. isoform-level DRIMSeq/stageR
Isoform-level testing only when replicates and depth support statistical power. Otherwise gene-level DE is reported, with the isoform catalog as descriptive output [10].

Common questions

What is the minimum RNA input, RIN, and sequencing depth for long-read RNA-seq analysis?

Pepkio recommends ≥300 ng total RNA with RIN >7 and DNase treatment for PacBio Kinnex Iso-Seq libraries [2, 3]. PacBio recommends 5 million HiFi reads per sample for high-abundance isoform discovery and 10 million for moderate-to-rare isoforms or isoform-level differential expression [2]. ONT depth is scaled using SG-NEx protocol benchmarks [6]. ≥3 biological replicates per condition are recommended for DE. Exact targets are confirmed at kickoff.

Can you analyze low-yield or degraded RNA samples?

Yes, with limitations documented in the QC report. Samples below RIN 7 often produce truncated reads that inflate artifact isoforms in uncured catalogs [1]. Pepkio flags elevated truncation and low full-length fractions before SQANTI3 filtering. Re-extraction or re-sequencing is discussed when yield cannot support the planned contrasts.

Do you support PacBio Revio Kinnex, ONT direct RNA, ONT cDNA, and MGI CycloneSEQ libraries?

Yes. PacBio Revio and Sequel II Kinnex HiFi data are processed via isoseq3 (when needed) and minimap2 splice:hq. ONT direct RNA, amplification-free cDNA, and PCR-cDNA libraries use minimap2 splice with ONT-appropriate parameters [6, 7, 8]. MGI CycloneSEQ-WT02/WY01 FASTQs use minimap2 splice when scoped at kickoff [13]. Library-type-specific QC metrics are reported in the deliverable bundle.

How long does long-read RNA-seq analysis take at Pepkio?

A standard bulk cohort (roughly 6-12 samples, reference-guided, one contrast) typically completes in 4-6 weeks from data receipt. Annotation-free discovery, multi-condition isoform DE, or orthogonal short-read junction validation may take 6-10 weeks. Exact timelines are confirmed at kickoff.

How do you handle batch effects across sequencing runs or library prep batches?

For gene-level DESeq2, batch is included as a covariate in the design matrix when the same biological contrast spans multiple sequencing runs [12]. Isoform-level DRIMSeq models include batch when replicate structure supports it. After fitting batch covariates, Pepkio inspects top DE genes and marker isoforms to check that condition-associated signal is still present before final delivery.

Do I own the code, and in what format is it delivered?

Yes. You retain full ownership of all code, scripts, and results. Pepkio delivers commented R/Python scripts and environment lock files (sessionInfo(), conda, or pip). Count matrices and GTF files use standard formats readable in IGV, tappAS, or Bioconductor. R Markdown or Jupyter delivery is available on request.

Can I be involved during analysis?

Yes. Checkpoint reviews occur after read QC, after SQANTI3 filtering, and before final delivery. You can review structural-category distributions, adjust filter stringency within agreed scope, and request additional contrasts. A PhD-level scientific contact leads the project and incorporates your tissue-specific knowledge.

What does post-delivery reviewer support include?

Clarification of computational methods, SQANTI3 filter thresholds, and minor figure or table revisions within agreed scope (typically ≤20% of deliverables). Pepkio drafts Methods and Supplementary text for analyses we performed. Substantial new analyses requested by reviewers are scoped separately.

Is co-authorship required?

No. Pepkio operates as a fee-for-service provider and does not require co-authorship unless that is discussed in advance. Acknowledgment of bioinformatics support in the Acknowledgments section is the usual practice. Co-authorship is considered only when Pepkio scientists make substantial intellectual contributions beyond routine analysis.

How many reads do I need to discover novel isoforms?

PacBio recommends 5 million HiFi reads per sample for high-abundance isoform discovery and 10 million for moderate-to-rare transcripts on Kinnex libraries [2]. Downsampling of heart and brain Kinnex data suggests ~80% of known genes and isoforms may be detectable at 10-20 million reads per sample ([2], ESHG poster). LRGASP advises additional replicates and orthogonal data when the goal is rare or novel transcript detection [5].

How do you produce count matrices and validate novel isoforms after SQANTI3 filtering?

SQANTI3 read counts and TPM values are for structural QC only, not for statistical testing (ConesaLab SQANTI3 wiki; [4]). Pepkio re-runs IsoQuant against the SQANTI3-filtered GTF per sample to produce DESeq2- and DRIMSeq-ready count matrices. When matched short-read BAMs are available, SQANTI3 --short_reads input adds junction validation metrics [4]. RT-PCR primer design for experimental validation is scoped separately.

Can you handle custom or non-standard long-read RNA-seq analyses?

Yes. Beyond the standard FASTQ-to-filtered-GTF workflow, Pepkio scopes bespoke work at kickoff: custom inputs (pre-aligned BAM, client GTF), fusion prioritization, ORF prediction, allele-specific isoform phasing, single-cell long-read (MAS-ISO-seq), or integration with short-read or spatial data [1]. Milestone pricing and timelines are confirmed before work begins.

Related services

  • Bulk RNA-seq — Gene-level differential expression when isoform resolution is unnecessary.
  • Single-cell RNA-seq — Cell-type heterogeneity without native full-length isoform capture.
  • Spatial transcriptomics — Tissue architecture and region-specific splicing context.
  • Whole-genome sequencing — Genomic variants and structural events that can affect splicing and transcript structure.
  • Custom consulting — Pre-sequencing depth, Kinnex multiplexing, and replicate design before library prep.
References
  1. Amarasinghe SL, Su S, Dong X, et al. Opportunities and challenges in long-read sequencing data analysis. Genome Biology. 2020;21(1):30. https://doi.org/10.1186/s13059-020-1935-5 (PMID: 32033565)
  2. Pacific Biosciences. Application note: Kinnex full-length RNA kit for isoform sequencing. 2024. https://www.pacb.com/wp-content/uploads/Application-note-Kinnex-full-length-RNA-kit-for-isoform-sequencing.pdf
  3. UCSD gCore Genomics Core. PacBio Revio bulk RNA Iso-Seq sample submission guidelines. 2024. https://gcore.ucsd.edu/RevioSub
  4. Pardo-Palacios FJ, Arzalluz-Luque Á, Kondratova L, et al. SQANTI3: curation of long-read transcriptomes for accurate identification of known and novel isoforms. Nature Methods. 2024;21(5):793–797. https://doi.org/10.1038/s41592-024-02229-2 (PMID: 38509328)
  5. Pardo-Palacios FJ, Wang D, Reese F, et al.; LRGASP Consortium. Systematic assessment of long-read RNA-seq methods for transcript identification and quantification. Nature Methods. 2024;21(7):1349–1363. https://doi.org/10.1038/s41592-024-02298-3 (PMID: 38849569)
  6. Chen Y, Davidson NM, Wan YK, et al. A systematic benchmark of Nanopore long-read RNA sequencing for transcript-level analysis in human cell lines. Nature Methods. 2025;22(4):801–812. https://doi.org/10.1038/s41592-025-02623-4 (PMID: 40082608)
  7. Prjibelski AD, Mikheenko A, Joglekar A, et al. Accurate isoform discovery with IsoQuant using long reads. Nature Biotechnology. 2023;41(7):915–918. https://doi.org/10.1038/s41587-022-01565-y (PMID: 36593406)
  8. Li H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics. 2018;34(18):3094–3100. https://doi.org/10.1093/bioinformatics/bty191 (PMID: 29750242)
  9. Aguzzoli Heberle B, Li H, Pardo-Palacios FJ, et al. Mapping medically relevant RNA isoform diversity in the aged human frontal cortex with deep long-read RNA-seq. Nature Biotechnology. 2024;42(11):1614–1622. https://doi.org/10.1038/s41587-024-02245-9 (PMID: 38778214)
  10. Du MRM, Gouil Q, Kawaji H, et al. Benchmarking long-read RNA-sequencing analysis tools using in silico mixtures. Nature Methods. 2023;20(12):1810–1821. https://doi.org/10.1038/s41592-023-02026-3 (PMID: 37783886)
  11. Nowicka A, Robinson MD. DRIMSeq: a Dirichlet-multinomial framework for multivariate count outcomes in genomics. F1000Research. 2016;5:1356. https://doi.org/10.12688/f1000research.8930.1
  12. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology. 2014;15(12):550. https://doi.org/10.1186/s13059-014-0550-8 (PMID: 25516281)
  13. MGI Tech. CycloneSEQ technology. 2024. https://global-mgitech.com/technologies/cycloneseq-technology/

Let's Talk About Your Science

Tell us:

  • • Your biological question
  • • Data type and size
  • • Timeline constraints

We'll tell you:

  • • What's feasible
  • • How long it will take
  • • Exactly what it will cost
Discuss Your Project

Contact us to start with a free consultation. Need everyday bench calculators? Try our free lab tools.