
Genomic artificial intelligence (AI) is moving from task-specific models toward systems that can integrate genomic, molecular, and phenotypic information across increasingly complex analytical workflows. There are governance frameworks for the responsible use of genomic AI, largely developed for systems whose data use and analytical behavior are defined in advance. Agentic systems add a new governance challenge because analytical decisions, resource use, and subsequent actions may no longer be fully specified before execution. In this review, we examine how this shift changes the governance of genomic AI and suggest that oversight must extend to the behavior of the analytical system itself. Permissions, data use, provenance, validation, uncertainty, and human authority need to remain enforceable during execution. We introduce AI systems implementation at the Center for Applied Genomics (CAG) of the Children’s Hospital of Philadelphia (CHOP), including current controls, an ongoing experiment using a restricted project-specific environment for agentic analysis, and a phased approach to broader deployment. Agentic genomic AI governance requires a shift from static review of data and models to enforceable control of system behavior during execution.
BackgroundFatigue represents a common and disabling extra-articular manifestation of rheumatoid arthritis (RA). However, the extent to which chronic inflammatory diseases share causal genetic links with fatigue remains elusive. This study was therefore designed to identify the causal drivers of fatigue among inflammatory conditions and to elucidate the underlying mechanisms.MethodsFourteen chronic inflammatory diseases (CIDs) were screened using multi-sample and two-sample Mendelian randomization (MR). Key genes were subsequently identified through an integrative pipeline comprising colocalization, transcriptome analysis, single-cell RNA sequencing, in silico knockout, and clinical validation. To quantify mediating pathways, mediation MR analysis was conducted. Finally, drug target prediction was performed using the Comparative Toxicogenomics Database (CTD) and molecular docking.ResultsThe pathogenic effect of RA on fatigue appears to be mediated primarily by anemia rather than direct genetic correlation. ASCC2 interacts with MMP3 and may represents a potential therapeutic target for folic acid and valproic acid. Furthermore, MMP3, IL-1β, and TNF-α were progressively elevated from NC to RA without fatigue (RA group) to RA with fatigue (RF group), whereas ASCC2 was upregulated in the RA group but declined modestly in the RF group.ConclusionRA may represent a unique causal contributor to fatigue, with its effects being potentially mediated primarily by anemia. Furthermore, ASCC2 appears to be a potential therapeutic target for RA-related fatigue.
Short-read exome and genome sequencing have transformed not only clinical genetics but also many other medical specialties that rely on genomic information. Yet, in clinical genetics, many individuals remain undiagnosed because the relevant disease mechanism cannot always be represented as an isolated variant at a single genomic coordinate. Phase, structural configuration, repeat composition, paralog assignment, methylation, and transcript consequences can each influence whether a variant is pathogenic. Long-read sequencing preserves molecular continuity across kilobases of native DNA and, in separate, complementary assays, across full-length RNA transcripts. Its clinical value includes both detecting previously missed variation and reconstructing the molecular configuration of detected findings. In this Perspective, we argue that clinical genomics should increasingly treat haplotypes, structural alleles, repeat alleles, and allele-specific regulatory states as primary diagnostic objects. Using a composite unresolved recessive-disease scenario, we place this perspective within the diagnostic odyssey, discuss health economics and outcomes research (HEOR) considerations, propose a context-aware workflow, and define priorities for validation and evidence generation. We identify potential value under reanalysis as a hypothesis requiring economic evaluation and argue that most current clinical databases are not designed to adequately represent the configurations these assays produce. Long-read sequencing need not universally replace short-read testing; its most compelling role is in settings where preserving molecular continuity changes the clinical meaning and downstream value of the result.
BackgroundFatal cardiomyopathy in infancy, particularly when accompanied by metabolic abnormalities and multiorgan failure, may result from pathogenic variation in either the nuclear or mitochondrial genome. Establishing causality remains challenging when candidate variants are of uncertain significance, show incomplete segregation, or are also present in clinically unaffected relatives.MethodsFamily-based whole-genome sequencing was performed on blood-derived DNA from the affected infant and both parents and on myocardial DNA from the infant. Primary variant analysis used a GATK-based pipeline with HaplotypeCaller and VEP annotation, followed by clinical interpretation and prioritization with VarSome Clinical and Franklin (Genoox) according to ACMG/AMP criteria, population frequency, predicted effect, biological relevance, inheritance, and family segregation. Structural variants were detected and analysed with DELLY v0.9.1. For computational confirmation of the mitochondrial findings, all four WGS datasets were reanalysed in Galaxy Europe using GATK 4.6.2.0 Mutect2 in mitochondrial mode, followed by HaploGrep3 haplogroup assignment.ResultsThe infant developed rapidly progressive cardiomyopathy with severe systolic dysfunction and multiorgan failure and died at 8 months of age. Fractional shortening decreased from 26% to 10%, and a large circumferential pericardial effusion required drainage of approximately 150 mL. Urinary lactate reached 2,192 mmol/mol creatinine, followed by hepatic and renal failure. Myocardial histology demonstrated diffuse interstitial fibrosis, extensive loss of myofibrillar structures, cytoplasmic vacuolization, and inclusion bodies, without evidence of myocarditis. Nuclear WGS identified candidate variants in FOXO6, VWF, and TSC1, but segregation and phenotype assessment did not support any as the primary disease cause. DELLY identified 27,214 structural-variant records; after stringent quality, segregation, genotype, and read-support filtering, four constitutional de novo and six myocardium-associated candidates remained, but none was considered causal without further validation. Dedicated mitochondrial-mode reanalysis reproduced near-homoplasmic m.4401A>G in proband blood (99.8%; 2,056×), myocardium (99.9%; 2,052×), and maternal blood (99.9%; 2,030×), with no raw Mutect2 call in paternal blood. The maternal-lineage variants m.772A>G and m.16288T>C showed a similar segregation pattern. HaploGrep3 assigned proband blood, myocardium, and maternal blood to haplogroup J1d3a, whereas paternal blood was assigned to N1b1a. The m.4401A>G variant was classified as a global private mutation relative to J1d3a rather than a haplogroup-defining variant and showed no enrichment in affected myocardial tissue.ConclusionNeither nuclear-genome nor mitochondrial-genome analyses identified a variant that could be confidently established as the cause of the observed phenotype. The mitochondrial variant m.4401A>G remains of biological interest but currently lacks sufficient genetic evidence for pathogenicity in this family.
BackgroundNeuronal intranuclear inclusion disease (NIID) is a rare neurodegenerative disorder characterized by a broad spectrum of clinical manifestations, including progressive dementia, limb weakness, cerebellar ataxia, and autonomic dysfunction. Peripheral nerve involvement is a recognized clinical feature of NIID and may be prominent in patients with NOTCH2NLC GGC repeat expansions.Case presentationWe report a familial case of NIID with prominent peripheral nerve involvement in a 48-year-old woman. The patient developed asymmetric limb weakness, sensory deficits on neurological examination, and postural tremor without the characteristic corticomedullary junction hyperintensity on diffusion-weighted imaging. Myasthenia gravis and anxiety-related episodes were initially considered during the diagnostic evaluation, but subsequent investigations supported a diagnosis of NIID. Nerve conduction studies revealed reduced conduction velocities in multiple motor and sensory nerves, raising suspicion of NIID. The diagnosis was confirmed by identification of a NOTCH2NLC GGC repeat expansion and skin biopsy showing intranuclear inclusions. Family screening identified additional NOTCH2NLC GGC repeat-expansion carriers, including the proband’s younger brother, who showed electrophysiological abnormalities suggestive of subclinical peripheral nerve involvement despite the absence of overt neurological symptoms.ConclusionNIID may exhibit substantial phenotypic heterogeneity and may present with prominent peripheral nerve involvement even in the absence of characteristic corticomedullary DWI abnormalities. Electrophysiological abnormalities may also be detected in clinically asymptomatic expansion carriers. Although this observation is based on a single family and has limited representativeness, it provides a rationale for further evaluation in larger cohorts.