Direct RNA Sequencing: Native Biology Without the Loss

Information Lost in cDNA Conversion

Traditional RNA-to-cDNA sequencing loses three critical data types: RNA base modifications (m6A, pseudouridine), poly(A) tail length information (which reveals transcript lifecycle and stability), and complete isoform mapping due to incomplete reverse transcription and PCR bias. PCR amplification also skews expression quantitation, overamplifying short transcripts and failing on sequences with repeats.

Data and efficiency losses in traditional cDNA RNA sequencing

Direct RNA Preserves Native Biology

Direct RNA sequencing reads native RNA molecules without cDNA conversion, preserving RNA modifications, measuring exact poly(A) tail lengths, capturing complete isoforms, and eliminating PCR bias for true expression quantitation. Results are faster (real-time data off the machine) and more affordable due to scalability across MinION to PromethION platforms.

Data richness and efficiency gains in direct RNA sequencing

How Nanopore Sequencing Works

A protein nanopore embedded in a membrane allows RNA or DNA to translocate through while a motor protein controls speed. As the molecule passes, it blocks ionic current in characteristic patterns specific to each base. These current signatures are base-called using trained machine-learning models to produce sequence data. DNA moves at ~400 bases/second; RNA at ~100–125 bases/second.

Motor protein translocation rates for DNA vs. RNA

Read-Length Agnostic, PCR-Free

Nanopore sequencing produces reads as long as the input molecule—there is no theoretical length limit. Because no PCR amplification is used, the sample composition entering the flow cell is exactly what you sequence, eliminating amplification bias and enabling accurate quantitation of transcript abundance.

New RNA-Specific Pore and Motor Protein

The RNA4 chemistry introduces a new RNA-specific nanopore (replacing the dual-purpose R9 pore, now discontinued) and an improved motor protein optimized for RNA. These work only with direct RNA preps and RNA-specific flow cells (MinION for RNA, ProMethION for RNA), delivering higher accuracy and better output than the previous RNA002 kit.

Accuracy Improvements: 98.8% Super-Accuracy

Base-calling models have been retrained on expanded datasets including modified-base data. High-accuracy model now achieves 97.5% for native RNA with modifications; super-accuracy model reaches 98.8%—approaching DNA-level accuracy. Backward compatible: existing RNA4 data can be reprocessed with new models to improve accuracy.

Accuracy improvement in high-accuracy base-calling model

Output Boost: 20% Increase via Software

New run conditions (software changes only, no prep changes) released May 2024 improve pore occupancy, yielding ~20% more reads on PromethION flow cells after 14–20 hours of sequencing. Available in MinKNOW 6.0 (coming soon).

20%

Output increase from new run conditions

Improvement in PromethION flow cell yield via software optimization

Faster Prep: 20 Minutes Saved

A new reverse transcriptase in the protocol reduces library prep time by 20 minutes (from ~2 hours to ~100 minutes) while enabling capture of longer cDNA-RNA hybrids, improving full-length transcript recovery.

Library prep time reduction

Short-Read Detection: 50 Nucleotides and Up

Software update enables high-quality, mappable output for reads as short as 50 nucleotides (previously 200 nt minimum). This unlocks analysis of tRNAs (~70 nt), degraded samples, and sheared RNA. A 90-nucleotide in vitro transcript on PromethION yielded >100 million reads, demonstrating massive output for short-read applications.

Minimum read length threshold for high-quality output

RNA Degradation: The Critical Risk

RNases are ubiquitous; degradation destroys transcript integrity. Mitigation requires: clean gloves, RNase-free reagents (RNase Zap), careful handling, and optimized extraction protocols. Full-length transcripts are essential because each transcript matters in RNA-seq—you need complete isoforms to identify which variant is expressed.

RNA Integrity Number (RIN) Threshold

RIN is a 1–10 scale measuring degradation by comparing 28S and 18S ribosomal RNA peaks on a bioanalyzer. RIN ≥7 is recommended; RIN 6 shows some degradation; RIN 2–3 indicates severe degradation. Oxford Nanopore provides guidance on effects of lower RIN samples.

RNA Integrity Number (RIN) scale and recommendations

Poly(A) Enrichment vs. Total RNA

Total RNA is ~95% ribosomal and transfer RNA; mRNA is only 1–5%. Enrichment strategies include poly(A) selection (oligodT pull-down, captures mRNA and long non-coding RNA) or ribosomal depletion (removes rRNA and tRNA). Oxford Nanopore provides protocols for both. Direct RNA kit accepts 1 µg total RNA or 300 ng poly(A)-enriched RNA.

Composition of total RNA extract

Poly(A) Tailing for Non-Polyadenylated RNA

The direct RNA kit targets poly(A) tails. Non-polyadenylated RNA (e.g., some tRNAs) can be sequenced by adding a synthetic poly(A) tail using poly(A) polymerase. Optimal reaction time is 30–90 seconds, not 30 minutes as often assumed. Oxford Nanopore provides validated protocols.

Library Prep: Reverse Transcription Adapter + Optional cDNA Synthesis

A reverse transcription adapter (RTA) with poly(T) overhang ligates to the poly(A) tail at the 3' end. Optional reverse transcription generates an RNA-DNA hybrid (not full cDNA conversion), which removes secondary structure to help the motor protein translocate smoothly. The hybrid is never sequenced; only the RNA strand generates signal for base-calling.

Direct RNA library preparation workflow

Read Length: Longest Observed ~27–28 KB

Sequencing is read-length agnostic. In-house, the longest transcript was ~19 KB (in vitro transcribed). Externally, customers have sequenced the full-length coronavirus genome (~27–28 KB). Read length depends entirely on input RNA quality and size.

27–28 KB

Longest observed read (coronavirus genome)

Maximum read length achieved in direct RNA sequencing

High Correlation Between Replicates

Technical replicates of Lexogen spike-in mixes show 99% correlation in gene counts and transcript-length representation. Biological replicates of human universal reference RNA (UHRR) show 98–99% correlation (Pearson/Spearman). This confirms robust, reproducible data generation.

99%

Correlation between technical replicates

Gene count correlation across replicate libraries

Output Metrics: PromethION vs. GridION

PromethION flow cells yield ~20 million reads; GridION/MinION flow cells yield ~5 million reads (fewer pores). Median read length for human transcriptome is ~1 KB. Majority of transcripts show ≥100% full-length coverage, confirming complete isoform capture.

Typical sequencing output by platform

Gene Expression & Isoform Mapping

Direct RNA enables robust measurement of transcript abundance across dynamic ranges (low to high expression). Full-length reads allow identification of all isoforms from a single gene, revealing which splice variants are used in specific cell types or conditions. This is critical for understanding gene regulation, disease mechanisms, and identifying drug targets.

Long Non-Coding RNA Discovery

Long non-coding RNAs (lncRNAs) are often >10 KB and difficult to sequence with short-read or cDNA methods. Direct RNA captures them in full, enabling comprehensive annotation of lncRNA function in development, disease, and synthetic biology.

Poly(A) Tail Length as a Biomarker

Poly(A) tail length reflects transcript stability and translational status. Knockouts or perturbations may shorten tails, reducing translation. Tail length differs between isoforms of the same gene and between mRNA and lncRNA. Dorado now outputs poly(A) tail estimates automatically, enabling studies of transcript lifecycle and disease states (e.g., cancer metastasis).

RNA Modifications: m6A and Pseudouridine

Direct RNA detects native RNA modifications without loss. Two m6A models are available: DRACH-motif (high accuracy, specific) and all-context (broader detection). A pseudouridine (Ψ) model identifies methylated uridines. These modifications regulate translation, stability, and immune response. Modifications are especially important in mRNA therapeutics and vaccine development.

mRNA Therapeutic & Vaccine QC

Direct RNA sequencing consolidates multiple QC tests (identity, integrity, purity, poly(A) tail length, modified-base detection) into a single assay on one platform. A specialized model detects N1-methyl-pseudouridine (used in many mRNA vaccines) at 97.2% accuracy. This simplifies manufacturing QC and reduces time, cost, and footprint compared to traditional multi-method approaches.

Critical quality attributes testable in single direct RNA run

Multiplexing (In Development)

Oxford Nanopore is actively developing multiplexing for direct RNA sequencing to enable simultaneous sequencing of multiple samples on a single flow cell. Release timeline is 'sooner rather than later,' but no specific date has been announced.

Targeted RNA Sequencing

Targeted sequencing will allow enrichment of specific transcripts or regions of interest, reducing sequencing cost per target and enabling deep coverage of genes of interest.

Automation & Integration

Automation is being developed for Elution systems and third-party liquid handlers, reducing manual labor and improving throughput for high-volume labs.

tRNA Sequencing Protocols

Oxford Nanopore is developing and validating protocols for tRNA sequencing, including methods for non-polyadenylated tRNAs and detection of tRNA modifications.