Diagnosing genetic disorders requires extensive manual curation and interpretation of candidate variants, a labor-intensive task even for trained geneticists. Although artificial intelligence (AI) shows promise in aiding these diagnoses, existing AI tools have only achieved moderate success for primary diagnosis. AI-MARRVEL (AIM) uses a random-forest machine-learning classifier trained on over 3.5 million variants from thousands of diagnosed cases. AIM additionally incorporates expert-engineered features into training to recapitulate the intricate decision-making processes in molecular diagnosis. The online version of AIM is available at https://ai.marrvel.org. To evaluate AIM, we benchmarked it with diagnosed patients from three independent cohorts. AIM improved the rate of accurate genetic diagnosis, doubling the number of solved cases as compared with benchmarked methods, across three distinct real-world cohorts. To better identify diagnosable cases from the unsolved pools accumulated over time, we designed a confidence metric on which AIM achieved a precision rate of 98% and identified 57% of diagnosable cases out of a collection of 871 cases. Furthermore, AIM’s performance improved after being fine-tuned for targeted settings including recessive disorders and trio analysis. Finally, AIM demonstrated potential for novel disease gene discovery by correctly predicting two newly reported disease genes from the Undiagnosed Diseases Network. AIM achieved superior accuracy compared with existing methods for genetic diagnosis. We anticipate that this tool may aid in primary diagnosis, reanalysis of unsolved cases, and the discovery of novel disease genes. (Funded by the NIH Common Fund and others.)
Agammaglobulinemia is the most profound primary antibody deficiency that can occur due to an early termination of B-cell development. We here investigated 3 novel patients, including the first known adult, from unrelated families with agammaglobulinemia, recurrent infections, and hypertrophic cardiomyopathy (HCM). Two of them also presented with intermittent or severe chronic neutropenia. We identified homozygous or compound-heterozygous variants in the gene for folliculin interacting protein 1 (FNIP1), leading to loss of the FNIP1 protein. B-cell metabolism, including mitochondrial numbers and activity and phosphatidylinositol 3-kinase/AKT pathway, was impaired. These defects recapitulated the Fnip1−/− animal model. Moreover, we identified either uniparental disomy or copy-number variants (CNVs) in 2 patients, expanding the variant spectrum of this novel inborn error of immunity. The results indicate that FNIP1 deficiency can be caused by complex genetic mechanisms and support the clinical utility of exome sequencing and CNV analysis in patients with broad phenotypes, including agammaglobulinemia and HCM. FNIP1 deficiency is a novel inborn error of immunity characterized by early and severe B-cell development defect, agammaglobulinemia, variable neutropenia, and HCM. Our findings elucidate a functional and relevant role of FNIP1 in B-cell development and metabolism and potentially neutrophil activity.
The histone H3 variant H3.3, encoded by two genes H3-3A and H3-3B, can replace canonical isoforms H3.1 and H3.2. H3.3 is important in chromatin compaction, early embryonic development, and lineage commitment. The role of H3.3 in somatic cancers has been studied extensively, but its association with a congenital disorder has emerged just recently. Here we report eleven de novo missense variants and one de novo stop-loss variant in H3-3A (n = 6) and H3-3B (n = 6) from Baylor Genetics exome cohort (n = 11) and Matchmaker Exchange (n = 1), of which detailed phenotyping was conducted for 10 individuals (H3-3A = 4 and H3-3B = 6) that showed major phenotypes including global developmental delay, short stature, failure to thrive, dysmorphic facial features, structural brain abnormalities, hypotonia, and visual impairment. Three variant constructs (p.R129H, p.M121I, and p.I52N) showed significant decrease in protein expression, while one variant (p.R41C) accumulated at greater levels than wild-type control. One H3.3 variant construct (p.R129H) was found to have stronger interaction with the chaperone death domain-associated protein 6.
COPA syndrome, an autosomal-dominant inborn error of immunity, is nonpenetrant in ∼20% of individuals, with no known mediators of protection. Recent studies implicate STING in the pathogenesis of COPA syndrome. We show that the common HAQ STING allele mediates complete clinical protection. We sequenced 35 individuals with COPA mutations, 26 affected patients and 9 unaffected carriers, finding HAQ STING co-segregation with clinical nonpenetrance. Exome sequencing identified only the mutations comprising HAQ STING as variants shared by unaffected carriers and absent in patients. Experimentally, we found that HAQ STING acts dominantly to dampen COPA-dependent STING signaling. Expressing HAQ STING in patient cells rescued the molecular phenotype of COPA syndrome. Our study is the first report of a common and well-tolerated allele mediating complete clinical protection from a severe genetic disorder. Our findings redefine the diagnostic criteria for COPA syndrome, expose functional differences among STING alleles with broad scientific and clinical implications, and reveal a potential universal gene therapy approach for patients.