Blood-based RNA-Seq of 5412 individuals with rare disease identifies new candidate diagnoses in the National Genomic Research Library

This study demonstrates that blood-based RNA sequencing of 5,412 individuals with rare diseases in the National Genomic Research Library successfully identifies clinically relevant diagnostic candidates by detecting expression and splicing outliers across a diverse range of disorders.

The Gist

Imagine you are trying to solve a massive, intricate jigsaw puzzle. For years, doctors have been trying to diagnose rare diseases by looking at the "blueprint" of a person's body—their DNA (genome). This blueprint is written in a code of four letters (A, C, G, T). However, just because you have the blueprint doesn't mean you can see how the factory is actually running. Sometimes, the blueprint has a typo, but the real problem is that the factory is reading the instructions wrong, skipping a page, or printing too many copies of a specific part.

This paper is about a team of scientists who decided to stop just looking at the blueprint and started watching the factory in action. They used a new tool called RNA-Seq to look at the "working copies" of the instructions (transcripts) that cells are actually using.

Here is the story of what they found, explained simply:

1. The Big Experiment: Watching the Factory

The researchers looked at 5,412 people with rare, undiagnosed diseases. These people had already had their DNA sequenced, but for most, the doctors couldn't find the "smoking gun" (the specific genetic error causing the illness).

Instead of just looking at the static DNA, the scientists took a blood sample from each person. Think of DNA as the master recipe book kept in a safe, and RNA as the active cooking notes the chef is using right now in the kitchen. By reading the cooking notes (RNA), the scientists could see if the chef was skipping ingredients, adding too much salt, or burning a specific dish.

2. The "Outlier" Hunt

The scientists used two smart computer programs (named OUTRIDER and FRASER2) to act like quality control inspectors. They looked at the cooking notes of all 5,412 people and asked: "Is anyone doing something weird compared to everyone else?"

• The "Too Much/Too Little" Inspector (OUTRIDER): This looked for genes that were being produced in weird amounts. Maybe a gene that should be quiet was screaming, or a gene that should be loud was whispering.
• The "Glitchy Sentence" Inspector (FRASER2): This looked for genes where the instructions were being cut and pasted incorrectly. Imagine a sentence that should read "The cat sat on the mat" but the printer skipped a word and printed "The cat sat on the."

The Result: They found these "glitches" in 20% of the people. That's 1 out of every 5 patients who had no diagnosis before now had a strong clue!

3. Why Blood? (The "Fruit Salad" Problem)

You might wonder: "Why look at blood? The disease might be in the brain or the muscles!"

This is a bit like trying to figure out how a car engine works by looking at the oil in the dipstick. It's not perfect. Some genes are only active in the brain or muscles, so they don't show up in the blood.

• The Good News: For many diseases (like those affecting the brain or development), the blood does show the problem.
• The Bad News: For some specific diseases (like certain muscle disorders), the blood is like a fruit salad where the "muscle fruit" is missing. The scientists found that for some disease panels, the blood samples were very good at showing the problem, but for others, they were "empty."

Despite this, using blood is a huge win because it's easy, safe, and cheap (just a needle prick) compared to taking a biopsy from a muscle or brain.

4. Catching the "Invisible" Criminals

The most exciting part of this paper is how RNA-Seq caught criminals that DNA sequencing missed. Here are a few examples of the "tricks" the DNA blueprint tried to hide:

• The Missing Page (Deletions): In one case, a tiny piece of the DNA was missing. The DNA test was too blurry to see it, but the RNA test showed that a whole "chapter" of the recipe was being skipped. It was like a printer missing a page, which the DNA scanner didn't notice, but the RNA scanner saw the gap immediately.
• The Wrong Cut (Splicing Errors): Sometimes the DNA looks fine, but the instructions tell the cell to cut the sentence in the wrong place. The RNA test showed the cell was cutting the sentence in the middle of a word, ruining the meaning.
• The Hidden Glitch (Deep Introns): Some errors happen in the "junk" parts of the DNA that usually get ignored. The DNA test often ignores these areas. But the RNA test showed that the cell was accidentally reading this "junk" and turning it into a harmful instruction.

5. The Verdict: A New Diagnostic Superpower

Before this study, many patients were stuck in "diagnostic limbo"—they knew they were sick, but doctors didn't know why.

This research shows that adding RNA-Seq to the standard DNA test is like giving the doctors a flashlight in a dark room.

• They found 78 new likely diagnoses just by looking at genes known to cause disease when one copy is broken.
• They found 39 more candidates by combining the RNA data with other clues.
• They even found cases where a patient had a "Variant of Uncertain Significance" (a genetic change that doctors weren't sure was bad). The RNA test proved, "Yes, this change is definitely breaking the machine," turning a mystery into a diagnosis.

The Bottom Line

This paper is a massive step forward. It proves that for a huge number of people with rare diseases, the answer isn't just in the DNA blueprint; it's in how that blueprint is being read. By looking at the "active notes" (RNA) in a simple blood test, doctors can finally solve mysteries that were previously impossible to crack, giving families answers and a path forward for treatment.

It's like realizing that to fix a broken car, you don't just need to read the manual; you need to listen to the engine running to hear exactly where the squeak is coming from.

Technical Summary

1. Problem Statement

Despite advances in whole-genome sequencing (WGS) and whole-exome sequencing (WES), more than 50% of patients with rare Mendelian disorders remain undiagnosed. Current diagnostic pipelines primarily focus on protein-coding variants, often overlooking:

• Splicing variants: Including those in non-canonical splice regions or deep intronic regions.
• Non-coding variants: Regulatory variants affecting gene expression.
• Structural Variants (SVs): Small deletions or complex rearrangements that may be filtered out by DNA-based callers due to low coverage or complexity.
• Tissue specificity: Many disease-relevant genes are not expressed in blood, limiting the utility of blood-based RNA-Seq, yet invasive tissue sampling (e.g., muscle biopsy) is not scalable.

The study aims to evaluate the scalability and diagnostic yield of blood-based total RNA-Seq as a complementary tool to WGS across a diverse, unselected cohort of rare disease patients.

2. Methodology

Cohort and Data Generation:

• Dataset: 5,412 individuals from Genomics England's 100,000 Genomes Project (100kGP) with diverse rare disorders.
• Sample Type: Whole-blood PAXgene samples (total RNA).
• Sequencing: Illumina NovaSeq 6000, 100bp paired-end reads, using a stranded total RNA prep kit with globin and rRNA depletion.
• Quality Control (QC): Strict thresholds applied (e.g., >120M total reads, >97% correct strand, >40X coding coverage). 5,412 samples passed QC.

Bioinformatic Pipeline:

• Alignment & Quantification: Performed using Illumina's DRAGEN pipeline (STAR aligner, Salmon quantification) against GRCh38/GENCODE v32.
• Outlier Detection:
  • Gene Expression: OUTRIDER (via the DROP pipeline) to identify expression outliers.
  • Splicing: FRASER2 (via DROP) to identify splicing outliers (deltaPSI ≥ 0.1 or ≤ -0.1).
• Prioritization Strategies:
  1. Haploinsufficiency (HI) Filtering: Focused on ClinGen HI genes (HI=3) within relevant PanelApp gene panels.
  2. Structural Variant (SV) Integration: Cross-referenced expression outliers with rare SVs (deletions, duplications, inversions) identified by Manta and Canvas.
  3. Phenotypic Integration: Used Exomiser to prioritize variants based on phenotype matching (HPO terms), SpliceAI scores (>0.2), and segregation.
  4. Manual Curation: Visual inspection of RNA-Seq and DNA reads in IGV to validate technical artifacts and confirm splicing mechanisms.

3. Key Contributions

• Scale: This is the largest cohort to date (n=5,412) utilizing blood-based RNA-Seq for rare disease diagnostics, moving beyond small, phenotype-specific studies.
• Scalability of Blood-Based RNA-Seq: Demonstrated that blood is a viable tissue source for a broad range of disorders, though capture efficiency varies significantly by disease type (e.g., high capture for congenital disorders of glycosylation vs. low for monogenic hearing loss).
• Diagnostic Yield: Identified at least one significant outlier event in a disorder-relevant gene in 20% of the cohort.
• Multi-Omics Integration: Successfully integrated RNA-Seq data with WGS, SV calling, and long-read sequencing (Nanopore) to resolve complex variants (e.g., retrotranspositions, small deletions missed by short-read DNA).
• VUS Resolution: Provided functional evidence to reclassify Variants of Uncertain Significance (VUS) to Likely Pathogenic by demonstrating aberrant splicing or expression.

4. Key Results

Cohort Characteristics:

• Median age at recruitment: 27 years.
• Primary disease groups: Neurology/neurodevelopmental (35%), Cardiovascular (15%), Renal (11%).
• Gene Capture: ~60% of PanelApp "green" (high-confidence) genes had median expression ≥5 TPM in blood. However, capture rates varied widely by panel (32% for congenital myopathy to 92% for congenital disorders of glycosylation).

Outlier Detection:

• Total Events: 36,239 expression outliers and 42,258 splicing outliers detected across the cohort.
• Relevant Events: In genes relevant to the patient's specific gene panel, 1,376 expression and 1,651 splicing outliers were found.
• Diagnostic Candidates: 20% of the cohort had at least one outlier in a relevant gene.
• Validation: Of 368 individuals with prior DNA-only diagnoses, RNA-Seq confirmed splicing disruptions in 47 cases; FRASER2 detected 48.9% of these at transcriptome-wide significance.

Case Studies & Mechanisms Identified:
The study presents 11 detailed case studies illustrating diverse mechanisms:

1. Small Deletions: A 1.55kb deletion in CTNNB1 (exon 14) was missed by DNA SV callers but detected as an exon-skipping event by RNA-Seq.
2. Deep Intronic Variants: Variants >50bp from splice sites (e.g., WDR26, KMT2D, SPAST, DNMT3A) causing cryptic splice site activation or pseudoexon creation.
3. Structural Variants:
   • Inversions: A 764kb inversion disrupting the 5'-UTR of DYRK1A and ANKRD11 led to reduced expression.
   • Retrotransposition: An SVA type D retrotransposition in APC intron 10 caused upregulation, resolved only via long-read sequencing.
4. Compound Heterozygosity: In INPPL1, a synonymous variant affecting splicing was identified alongside a frameshift, explaining the phenotype despite modest expression changes.
5. Expression vs. Splicing Discordance: Some loss-of-function variants resulted in overexpression (due to intron retention) or only modest expression changes (due to leaky splicing), highlighting the need to analyze both metrics.

Candidate Diagnoses:

• 78 candidates identified via HI gene splicing outliers.
• 19 candidates identified via SV-expression overlap.
• 39 candidate diagnoses identified via Exomiser integration (20 involving canonical splice variants, 19 involving non-canonical/deep intronic variants).

5. Significance

• Clinical Impact: The study demonstrates that adding blood-based RNA-Seq to the diagnostic workflow can resolve a significant portion of "unsolved" rare disease cases, particularly those involving non-coding variants, splicing errors, and complex structural variants.
• Cost-Effectiveness: It validates blood as a scalable, non-invasive alternative to tissue biopsies for a wide spectrum of disorders, despite variable gene expression capture.
• Pipeline Optimization: The findings suggest that diagnostic pipelines must move beyond canonical splice sites (±50bp) and integrate RNA-Seq data to interpret deep intronic variants and complex SVs.
• Resource Availability: The dataset and derived outlier tables are made available in the National Genomic Research Library (NGRL), serving as a foundational resource for future rare disease research and the discovery of novel disease genes.

In conclusion, this study establishes blood-based RNA-Seq as a critical, scalable component of the rare disease diagnostic toolkit, capable of uncovering pathogenic mechanisms that DNA sequencing alone frequently misses.