Specialist multidisciplinary input maximises rare disease diagnoses from whole genome sequencing

The diagnostic bottleneck in rare diseases has shifted from access to sequencing to interpretation of large genomic datasets. Clinical practice commonly applies virtual gene panels (VGPs) to filter WGS data, as in England’s 100,000 Genomes Project and National Health Service Genomic Medicine Service (NHS GMS), which pair centralised sequencing and bioinformatics with local clinical interpretation. Mainstream clinicians often receive WGS reports as definitive, but complexity arises due to missed variants when approaches are too narrow, variants of uncertain significance (VUS), and the value of post-WGS investigations such as reverse phenotyping and functional validation. These complexities motivate specialist multidisciplinary team (MDT) input for challenging rare disease cases. Primary mitochondrial diseases (PMDs), with broad, overlapping phenotypes across specialties, exemplify the need for integrated genomic medicine support. This study proposes and evaluates a clinician- and bioinformatician-led re-analysis pathway after routine WGS to improve diagnostic yield and clinical utility in adults with suspected PMD.

The paper situates its work within the transition to genome-wide testing guided by phenotype-linked virtual gene panels and the establishment of the NHS GMS informed by the 100,000 Genomes Project. Prior work shows that routine pipelines may miss diagnoses without broader bioinformatic strategies and post hoc investigations (e.g., reverse phenotyping, functional assays). Researcher-identified potential diagnoses can improve yield but are inconsistent due to variable access to clinical data and resources, and clinical scientists face workload limits that preclude extensive re-analyses. The authors argue for dedicated specialist MDTs to systematically address unsolved WGS cases. PMDs are highlighted as an exemplar due to phenotypic heterogeneity. The study contrasts centralized and private healthcare models and underscores the importance of specialized hub-and-spoke services integrating clinical genetics expertise.

Cohort and initial analyses: 102 adult patients (from 69 families) with suspected primary mitochondrial disease (PMD) underwent WGS after routine exclusion of common mtDNA and nuclear PMD causes. Ages ranged 17–81 years (mean 47.3); most were singletons (41.1%), with trios (27.4%), duos (21.6%), and larger pedigrees (6.9%). WGS used Illumina TruSeq on HiSeq 2500. Phenotypes were encoded as HPO terms from clinical notes. Routine analysis applied PanelApp-based virtual gene panels (green diagnostic-grade, with amber/red as insufficient evidence) and semantic similarity prioritisation (Exomis). Variants were tiered and reassessed by clinical scientists per ACMG criteria. Expanded specialist MDT re-analysis: For all non-diagnostic cases, a genomic medicine clinician re-reviewed phenotypes (including refining HPO terms), pedigrees, and applied modified Nijmegen mitochondrial criteria. Re-analysis expanded to: (1) include amber/red genes relevant to the phenotype, (2) add missed but phenotype-appropriate panels, (3) re-evaluate VUSs with updated clinical and literature evidence, (4) search for trans variants when a single heterozygous candidate was found in a recessive gene, and (5) reconsider inheritance models based on family history/segregation. Structural variant callers (‘Manta’ and ‘Canvas’) were used to prioritize CNVs; bespoke scripts reviewed CNVs overlapping coding sequence. Customized pipelines interrogated mtDNA, and a variant caller (MuteC2) was used to identify heteroplasmic mtDNA variants. When new diagnoses emerged, revised HPO terms and pedigrees were fed back into Exomis to assess capture by automated methods. Functional and confirmatory studies: Variants were validated by Sanger sequencing, with clinical scientist classification where uncertain. Functional assays supported pathogenicity for select non-coding variants (e.g., MYH2 transcript reduction >99% by qPCR; RNA analyses for NSUN3; upregulated expression for a C2orf6 variant). Additional clinical investigations (reverse phenotyping), imaging, muscle biopsy histochemistry/immunohistochemistry, Western blot, and qRT-PCR were performed as indicated. Cell culture and RNA extraction followed standard protocols. All participants provided informed consent; analyses occurred within the secure 100,000 Genomes Project environment under approved ethics.

- Routine semi-automated WGS analysis yielded diagnoses in 16.7% (17/102) of individuals in this complex PMD-suspected cohort. - Specialist MDT-led re-analysis provided an additional 14.7% (15/102) confirmed diagnoses, increasing the overall diagnostic rate to 31.4%; an additional 3.9% (4/102) had strong candidate (suspicious) VUSs in known or emerging disease genes. - Diagnostic categories among routine diagnoses included PMD (6/17, 35.3%), non-mitochondrial neurogenetic/neurodevelopmental disorders, muscular dystrophies, and cardiomyopathies; some cases partially explained the phenotype, suggesting potential dual diagnoses. - Contributors to new diagnoses included: • Detection of intronic variants in recessive genes (e.g., MOLCN1, POLR3, MYH2) missed initially; interpretation was aided by phenotypic correlation, with limited predictive support (only MYH2 showed a high SpliceAI score ~0.99). • Functional validation (e.g., MYH2 non-coding variant with >99% transcript reduction; RNA-based assessments for other candidates) to confirm pathogenicity. • Application of additional or updated virtual panels and inclusion of amber/red genes (e.g., POLR3, COL4A2, KIF22, CAPN1; CORO1A, NSUN7). • Pedigree interrogation and segregation analysis prompting revised inheritance assumptions and diagnoses in multiple families. • Reverse phenotyping (targeted clinical re-evaluation, imaging, skeletal survey, and muscle biopsy) supporting variant relevance in several cases. • Expanded filtering strategies identified heteroplasmic mtDNA variants and a mosaic variant in DM2. • Updated literature enabled reinterpretation/upgrading of variants (e.g., broadened phenotypic spectrum for KARS). - Clinical impact: All newly confirmed diagnoses had management implications, including eligibility for clinical trials or targeted therapies, adjustment of surveillance intensity, family counseling, and prioritization of system-specific screening.

The study demonstrates that specialist MDT input—combining genomic medicine clinicians, bioinformaticians, clinical scientists, and translational researchers—substantially improves diagnostic yield beyond routine semi-automated WGS pipelines in complex rare disease cohorts such as suspected PMD. Key drivers include refined phenotyping, targeted re-analysis with broader yet phenotype-appropriate panels (including amber/red genes), segregation assessment, and selective functional validation. These findings support embedding genomic medicine expertise within diagnostic services and adopting hub-and-spoke models to maximise clinical utility and equity of access. Purely annotation-driven approaches remain insufficiently sensitive for challenging variant classes (e.g., intronic, mosaic, heteroplasmic mtDNA) and for dual/complex phenotypes. Compared with researcher-led models, a structured MDT pathway offers a systematic and equitable solution that aligns with clinical care pathways, improves patient management, and informs counseling.

A specialist genomic MDT model nearly doubled the diagnostic rate for adults with suspected PMD after routine WGS, with additional high-confidence candidate findings and clear management benefits. The authors propose an evolved MDT workflow wherein post-routine re-evaluation refines phenotype and analysis parameters, feeds promising variants back to the diagnostic laboratory, and triggers targeted functional studies to secure pathogenicity. This approach offers a scalable, standardised framework to enhance diagnostic equity and clinical outcomes across rare diseases. Future directions include broader integration of RNA-based diagnostics, continual updating of VGP content, refined detection of complex variant classes (intronic, mosaic, heteroplasmic), and wider adoption of specialist MDT models within national genomic services.

Several challenges constrained interpretation and generalizability: (1) Intronic/non-coding variant interpretation was difficult without RNA data, with splicing prediction tools variably informative. (2) The cohort was highly selected and complex, which may limit direct comparison to broader rare disease populations. (3) Routine pipelines missed heteroplasmic mtDNA and mosaic variants without expanded filtering. (4) Reliance on updated literature and functional assays for variant upgrading indicates that some diagnoses depend on resources not universally available. (5) Full affiliation and panel details are centralised and some processes occurred within a secure data environment, potentially impacting external reproducibility of all analytic steps.