DNA sequencing can identify a plausible genetic change without showing whether it alters the messages a cell actually uses. RNA sequencing can add that missing observation: how a gene is expressed, how its RNA is assembled and whether one gene copy contributes differently from the other.
The most useful question is specific: “Can RNA from this sample help resolve this genetic finding?” This guide focuses on RNA analysis in inherited rare-disease investigation. It does not cover the separate uses of RNA sequencing for cancer biomarkers or infectious-disease testing.
Three kinds of evidence RNA can add
Messenger RNA is produced from DNA and processed before it is used to make protein. During splicing, selected RNA segments are joined together. RNA analysis can reveal changes in that processing or in the amount of transcript available. Kremer and colleagues demonstrated the diagnostic use of expression, splicing and allele-specific observations in patient-derived fibroblasts.
| Observation | Plain-language meaning | Follow-up question |
|---|---|---|
| Altered expression | More or less RNA than expected in the comparison. | Could sample composition or another biological process explain the difference? |
| Altered splicing | RNA segments are joined differently, such as skipping an exon or including intronic sequence. | What transcript is produced, and what could it do to protein function? |
| Allele-specific imbalance | The two gene copies contribute unequal amounts of detectable RNA. | Can the imbalance be linked reliably to the candidate variant? |
These are observations to investigate. Their value depends on how securely the laboratory connects them to the DNA finding and the clinical question.
Choose the sample before interpreting the result
A gene may be readily measurable in one tissue and poorly represented in another. Its RNA processing can also differ between tissues. A convenient blood sample therefore cannot be assumed to answer every question about muscle or another affected organ.
The muscle transcriptome study by Cummings and colleagues illustrates the importance of disease-relevant tissue and suitable reference samples. The investigators also assessed sample quality and tissue composition, rather than interpreting every unusual RNA pattern as a genetic mechanism.
Before collection, ask the laboratory whether the relevant gene and transcript are assessable in the proposed sample and what its validated test can detect. A more invasive sample should be considered with the clinical team, including whether it is likely to answer a meaningful question.
Example 1: a suspected splice change becomes measurable
Fictional teaching example. DNA testing identifies an intronic variant near an exon. A prediction tool suggests an effect on splicing. RNA analysis then detects a transcript that skips that exon.
That observation makes the question more concrete, but the report still needs to describe the effect. Is the skipped transcript present above the comparison background? How much normally processed RNA remains? Does the altered transcript plausibly disrupt the relevant gene function?
The ClinGen SVI splicing recommendations explain how experimental RNA evidence can inform an appropriately weighted PVS1 assessment for loss-of-function mechanisms. They also caution against adding another evidence code for the same splice effect as if it were independent confirmation. An abnormal splice result is not an automatic upgrade to pathogenic.
For a teaching discussion, write two sentences: “The assay observed…” and “To establish clinical significance, we still need…”. Keeping those sentences separate makes an interpretation much easier to defend. Our ACMG/AMP guide explains the wider classification framework.
Example 2: a negative test that cannot settle the question
Fictional teaching example. A clinician asks whether a candidate variant changes a transcript important in muscle. Blood RNA sequencing reports no abnormal splicing, but the relevant exon has too few reads for a dependable assessment.
“No abnormality detected” and “the region was adequately assessed and appeared normal” are different statements. In this example, the useful conclusion is that this sample did not provide an informative answer for that transcript. It would be misleading to rewrite it as “the DNA variant has no effect.”
RNA degradation can also complicate interpretation. Cells may remove some transcripts carrying premature stop signals through nonsense-mediated decay. The amount of an abnormal transcript detected can therefore depend on its stability as well as how it was produced. The ClinGen loss-of-function guidance discusses this biological context.
At the next discussion, ask for the reason the result is inconclusive and whether a different assay, another sample or reanalysis of existing DNA data would answer the original question. There is no single follow-up test appropriate for every unresolved case.
Example 3: an outlier is a lead, not yet an explanation
Fictional teaching example. An analysis flags unusually low expression of one gene. The research team now has a promising lead. Its next task is to investigate whether the signal is technically sound and connected to the person’s DNA findings.
Transcriptome-directed investigations can work in this direction, using an RNA signal to focus subsequent genomic analysis. The primary study by Murdock and colleagues describes that approach. A candidate emerging from an outlier screen still needs an evidence trail before it becomes a reported molecular diagnosis.
A helpful journal-club question is: “Which additional result would make us change our mind?” If no observation could challenge the favored explanation, the discussion may have moved from testing a hypothesis to defending it.
How much does RNA sequencing improve diagnosis?
Published yields describe particular cohorts and workflows. They are not a personal probability of diagnosis or a ranking of sample types. Three foundational studies illustrate why the denominator matters:
| Study | Setting | Reported finding |
|---|---|---|
| Kremer et al., 2017 | Fibroblast analysis in suspected mitochondrial disease. | Diagnoses in 5 of 48 previously undiagnosed cases, approximately 10%. |
| Cummings et al., 2017 | A selected muscle-disorder cohort studied with muscle RNA. | 17 previously unsolved families diagnosed; the study reported a 35% rate in its undiagnosed subset. |
| Lee et al., 2020 | A broader rare-disease investigation integrating genomic and RNA data. | Seven additional diagnoses among 48 genome-negative cases undergoing RNA analysis, about 15%. |
In the Lee study, those seven diagnoses also raised the overall rate in the 100-person analysis cohort from 31% to 38%. The 15% subgroup yield and seven-percentage-point overall increase describe different denominators; neither replaces the other.
When comparing a new paper with these studies, record who entered the cohort, what testing had already been completed, which samples were available and what the investigators counted as a diagnosis. Our paper-comparison workspace provides a place to put those details side by side.
A practical reading checklist
- State the question. Was RNA used to investigate a particular variant or search for unexplained abnormalities?
- Identify the assay. Whole-transcriptome sequencing, a targeted RNA assay and an engineered experimental system do not offer identical information.
- Check assessability. Was the relevant transcript measurable with adequate quality in the tested material?
- Inspect the comparison. What reference samples or controls make the observation unusual?
- Separate finding from interpretation. What did the assay observe, and what reasoning connects that finding to the disease mechanism?
- Record the remaining uncertainty. What requires confirmation, another source of evidence or further clinical assessment?
For functional experiments more generally, ClinGen’s functional-evidence recommendations emphasize matching the assay to the disease mechanism and evaluating validation and controls. The existence of a published experiment alone does not settle its evidential strength.
Questions to take to the laboratory or genetics appointment
- What specific uncertainty could RNA testing resolve in this case?
- Can your assay assess the relevant gene and transcript in the available sample?
- If the result is negative, what would still remain untested?
- Would a research finding need confirmation before it changes the clinical report?
- How will the RNA evidence be combined with the other findings?
If you are following a variant through different databases, continue with our guide to conflicting ClinVar classifications and variant research worksheet. The aim is a more precise question and a traceable explanation of the evidence—not a diagnosis generated from an isolated RNA plot.
Two worked examples: what would change the interpretation?
Example 1: an unusual splice product
In this fictional case, a child has a DNA variant near an exon boundary. The gene is expressed in the sample tested. RNA reads show a transcript that skips the exon, and that transcript is uncommon in suitable comparison samples. This supports a possible splice effect, but it does not complete the diagnosis.
The laboratory asks whether the transcript is relevant to the condition, whether the change alters the reading frame, whether RNA degradation affects the measured proportion, and whether the effect can be linked to the variant-bearing allele. It also checks the established disease mechanism and inheritance. An unexpected transcript can be important evidence without automatically changing a variant of uncertain significance to pathogenic.
Example 2: a reassuring-looking blood result that answers little
A second fictional patient has a suspected muscle disorder. Blood RNA shows no convincing abnormality, but the gene and relevant isoform have very low coverage. “No abnormal splicing detected” in that sample cannot establish that splicing is normal in muscle. The honest conclusion is that the study was not sufficiently informative for that question.
Before considering another sample, the team weighs whether it is likely to resolve a clinically meaningful uncertainty, whether a validated targeted test is available, and the burden of obtaining tissue. A biopsy should not follow automatically from an uninformative blood experiment.
What can long-read RNA sequencing add?
Short RNA reads often describe local junctions. Longer reads can help connect several exons within the same transcript, so changes that appear separate may be recognised as part of one RNA molecule. This is a question about transcript structure, distinct from using long-read DNA sequencing to resolve genomic structure.
The STRIPE study, published in 2026, examined a targeted long-read RNA approach in 88 individuals across two rare-disease groups. For eight of 15 splice-site-region variants, the investigators observed more complex RNA-processing effects than a simple exon-skipping or cryptic-splice model. This is a selected study of a particular method, not a universal diagnostic-yield estimate. Read the original study.
A worksheet for an RNA result discussion
| Record | Question to resolve |
|---|---|
| Sample and transcript | Was the relevant gene isoform measurable in this tissue? |
| Observation | Was there altered splicing, expression or allele-specific output? |
| Comparison | Were controls appropriate for tissue, processing and analysis? |
| Attribution | Can the effect be connected to the candidate DNA variant? |
| Clinical meaning | Does the effect fit the gene’s disease mechanism and this person’s findings? |
| Next step | Does the evidence justify confirmation, reclassification, further research or no change? |
Keep the molecular observation separate from the final clinical conclusion. Use the variant worksheet to organise sources and the VUS guide to discuss what remains uncertain. A useful report states both what the experiment showed and which question it could not answer.