Diagnostic yield of long-read sequencing in short-read-negative patients: a systematic review and meta-analysis

Long-read sequencing (LRS) is increasingly used for rare disease patients undiagnosed by short-read sequencing (SRS), but its additional yield has not been synthesized at the individual-patient level or across both targeted and genome-wide approaches, and the phenotypes and rescue mechanisms driving added diagnoses remain mostly uncharacterized. We assessed the LRS contribution across rare diseases by performing a systematic literature search in PubMed, Scopus, Web of Science, EMBASE and Google Scholar (to 31 October 2025), including studies that applied LRS to patients with monogenic rare disease undiagnosed after SRS. Studies were categorized into genome-wide cohort studies (Tier 1), targeted studies (Tier 2), and case reports/series (Tier 3), and random-effects meta-analysis was performed for Tier 1 studies with prespecified subgroup analyses. Patient phenotypes were encoded as Human Phenotype Ontology terms and hierarchically aggregated to test phenotype-specific associations with rescue. Across 11 Tier 1 studies (538 patients), the pooled rescue rate combining definitive and possible diagnoses was 16.4% (95% CI: 6.9–29.0%); restricting to definitive rescue only, the pooled yield was 12.0% (95% CI: 3.3–25.3%). Across 15 Tier 2 targeted studies (165 SRS-negative patients), the rescue rate was 37.6%, reaching 100% in MUC1 -associated ADTKD (12/12) and Canavan disease (2/2 SRS-negative and 8/8 second-hit cases) but only 3.9% in NF1 (2/51). Across 119 second-hit cases (Tier 1 + Tier 2), 34.5% were rescued. Among SRS-negative patients in the primary meta-analysis cohort, yield declined as prior testing deepened (ES-only 23.5% vs ES+GS 6.8%). Structural variants accounted for ~49% of genome-wide rescues, though ~29% of rescues reflected reanalysis findings rather than LRS-specific capability. In exploratory phenotype analysis, spasticity, microcephaly, and intellectual disability were among the phenotypes most strongly associated with rescue, a signal that persisted within neurological cases. LRS provides a meaningful rescue for patients with rare diseases undiagnosed after SRS, primarily through structural variant detection. Re-evaluation of existing SRS data and prioritization of patients directly after negative ES may optimize diagnostic yield, and phenotype-informed prioritization is a promising direction that warrants prospective evaluation.