Background: Blind spots in short-read-based diagnosis

A proposal calls for integrating long-read genome sequencing, diploid genome assembly, pangenome references, and artificial intelligence (AI)-based variant interpretation. A Perspective article in Nature Genetics brings these approaches together under “near-perfect genome sequencing” (NPGS) and outlines a phased path toward adoption in medical genetics. The central idea is to advance both technologies that capture more complete genomic information and systems for interpreting the clinical significance of variants.

The authors start from the limitations of fragmented diagnostic workflows based on short-read sequencing. Difficult-to-analyze regions remain, including regions with highly similar sequences, tandem repeats, and segmental duplications. They also identify large or complex structural variants and base modifications that are difficult to detect with existing methods as blind spots. They point to limitations in phasing—determining which of the homologous chromosomes carries each of multiple variants.

The paper’s question goes beyond reading more sequence. It also asks whether how completely a genome has been characterized can serve as evidence for assessing the clinical significance of variants.

Key findings: Using genome completeness as evidence for variant interpretation

The paper centers on a proposed reorganization of the analytical framework rather than new experimental results. The authors position long-read sequencing, which reads long stretches of DNA, as one component of NPGS. They propose connecting it with diploid genome assembly, which separately reconstructs the sequences of the two homologous chromosomes; pangenome references that represent diverse genomes; and AI-based variant interpretation. The emphasis is on combining technologies rather than introducing long-read sequencing alone.

For interpretation, the authors propose a Bayesian framework that treats genome completeness itself as evidence for variant classification. The aim is to incorporate the completeness of the available genomic information into the assessment, linking this approach to the interpretation of variants of uncertain significance (VUS) found in clinical practice. A VUS is a variant whose association with disease is difficult to establish conclusively.

The supplied abstract does not provide specific formulas for this framework, validation cohorts, or VUS reclassification rates. Neither the accuracy of a particular AI model nor diagnostic yield compared with existing tests could be established, so this material alone does not show how much variant interpretation would improve in practice.

Significance and outlook: A single-test vision and conditions for clinical adoption

The authors identify postnatal diagnosis, prenatal diagnosis, and oncology as potential applications for NPGS and discuss a phased adoption roadmap toward diagnosis centered on a single test. Cancer is one of several application areas; the supplied material does not contain results testing treatment efficacy in a specific solid tumor.

Practical challenges to adoption include cost, computational demands, equity, and ethical considerations. Applying an integrated analytical framework in clinical care requires assessing operational conditions alongside technical performance. The abstract provides no detailed adoption timeline, costs, or patient care outcomes. The term “near-perfect” should not be taken to mean an error-free test or a guarantee of diagnosis for every genetic disorder.

This proposal outlines a direction for considering genomic information acquisition and clinical interpretation within a single diagnostic framework. Its specific clinical utility remains a matter for further validation. The evidence is limited to the supplied publication details and abstract.