Identifying digenic disease genes via machine learning in the Undiagnosed Diseases Network.

Rare diseases affect millions of people worldwide, and discovering their genetic causes is challenging. More than half of the individuals analyzed by the Undiagnosed Diseases Network (UDN) remain undiagnosed. The central hypothesis of this work is that many of these rare genetic disorders are caused by multiple variants in more than one gene. However, given the large number of variants in each individual genome, experimentally evaluating combinations of variants for potential to cause disease is currently infeasible. To address this challenge, we developed the digenic predictor (DiGePred), a random forest classifier for identifying candidate digenic disease gene pairs by features derived from biological networks, genomics, evolutionary history, and functional annotations. We trained the DiGePred classifier by using DIDA, the largest available database of known digenic-disease-causing gene pairs, and several sets of non-digenic gene pairs, including variant pairs derived from unaffected relatives of UDN individuals. DiGePred achieved high precision and recall in cross-validation and on a held-out test set (PR area under the curve > 77%), and we further demonstrate its utility by using digenic pairs from the recent literature. In contrast to other approaches, DiGePred also appropriately controls the number of false positives when applied in realistic clinical settings. Finally, to enable the rapid screening of variant gene pairs for digenic disease potential, we freely provide the predictions of DiGePred on all human gene pairs. Our work enables the discovery of genetic causes for rare non-monogenic diseases by providing a means to rapidly evaluate variant gene pairs for the potential to cause digenic disease.

Probable digenic inheritance of Diamond-Blackfan anemia.

A 26-year-old female proband with a clinical diagnosis and consistent phenotype of Diamond-Blackfan anemia (DBA, OMIM 105650) without an identified genotype was referred to the Undiagnosed Diseases Network. DBA is classically associated with monoallelic variants that have an autosomal-dominant or -recessive mode of inheritance. Intriguingly, her case was solved by a detection of a digenic interaction between non-allelic RPS19 and RPL27 variants. This was confirmed with a machine learning structural model, co-segregation analysis, and RNA sequencing. This is the first report of DBA caused by a digenic effect of two non-allelic variants demonstrated by machine learning structural model. This case suggests that atypical phenotypic presentations of DBA may be caused by digenic inheritance in some individuals. We also conclude that a machine learning structural model can be useful in detecting digenic models of possible interactions between products encoded by alleles of different genes inherited from non-affected carrier parents that can result in DBA with an unrealized 25% recurrence risk.

Undiagnosed Disease Network collaborative approach in diagnosing rare disease in a patient with a mosaic CACNA1D variant.

The Undiagnosed Disease Network (UDN) is comprised of clinical and research experts collaborating to diagnose rare disease. The UDN is funded by the National Institutes of Health and includes 12 different clinical sites (About Us, 2022). Here we highlight the success of collaborative efforts within the UDN Clinical Site at Vanderbilt University Medical Center (VUMC) in utilizing a cohort of experts in bioinformatics, structural biology, and genetics specialists in diagnosing rare disease. Our UDN team identified a de novo mosaic CACNA1D variant c.2299T>C in a 5-year-old female with a history of global developmental delay, dystonia, dyskinesis, and seizures. Using a collaborative multidisciplinary approach, our VUMC UDN team diagnosed the participant with Primary Aldosteronism, Seizures, and Neurologic abnormalities (PASNA) OMIM: 615474 due to a rare mosaic CACNA1D variant (O'Neill, 2013). Interestingly, this patient was mosaic, a phenotypic trait previously unreported in PASNA cases. This report highlights the importance of a multidisciplinary approach in diagnosing rare disease.

The Genetic Landscape of Familial Pulmonary Fibrosis.

Rationale and Objectives: Up to 20% of idiopathic interstitial lung disease is familial, referred to as familial pulmonary fibrosis (FPF). An integrated analysis of FPF genetic risk was performed by comprehensively evaluating for genetic rare variants (RVs) in a large cohort of FPF kindreds. Methods: Whole-exome sequencing and/or candidate gene sequencing from affected individuals in 569 FPF kindreds was performed, followed by cosegregation analysis in large kindreds, gene burden analysis, gene-based risk scoring, cell-type enrichment analysis, and coexpression network construction. Measurements and Main Results: It was found that 14.9-23.4% of genetic risk in kindreds could be explained by RVs in genes previously linked to FPF, predominantly telomere-related genes. New candidate genes were identified in a small number of families-including SYDE1, SERPINB8, GPR87, and NETO1-and tools were developed for evaluation and prioritization of RV-containing genes across kindreds. Several pathways were enriched for RV-containing genes in FPF, including focal adhesion and mitochondrial complex I assembly. By combining single-cell transcriptomics with prioritized candidate genes, expression of RV-containing genes was discovered to be enriched in smooth muscle cells, type II alveolar epithelial cells, and endothelial cells. Conclusions: In the most comprehensive FPF genetic study to date, the prevalence of RVs in known FPF-related genes was defined, and new candidate genes and pathways relevant to FPF were identified. However, new RV-containing genes shared across multiple kindreds were not identified, thereby suggesting that heterogeneous genetic variants involving a variety of genes and pathways mediate genetic risk in most FPF kindreds.

A medical odyssey of a 72-year-old man with Charcot-Marie-Tooth disease type 2 newly diagnosed with biallelic variants in SORD gene causing sorbitol dehydrogenase deficiency.

A 72-year-old man was referred to the Undiagnosed Diseases Network (UDN) because of gradual progressive weakness in both lower extremities for the past 45 years. He was initially diagnosed as having Charcot-Marie-Tooth disease type 2 (CMT2) without a defined molecular genetic cause. Exome sequencing (ES) failed to detect deleterious neuromuscular variants. Very recently, biallelic variants in sorbitol dehydrogenase (SORD) were discovered to be a novel cause of inherited neuropathies including CMT2 or distal hereditary motor neuropathy (dHMN) referred to as Sorbitol Dehydrogenase Deficiency with Peripheral Neuropathy (SORDD, OMIM 618912). The most common variant identified was c.757delG; p.A253Qfs*27. Through the Vanderbilt UDN clinical site, this patient was formally diagnosed with SORDD after the identification of homozygosity for the above SORD frameshift through UDN Genome Sequencing (GS). His medical odyssey was solved by GS and detection of extremely high levels of sorbitol. The diagnosis provided him the opportunity to receive potential treatment with an investigational drug in a clinical trial for SORDD. We suggest that similar studies be considered in other individuals thought to possibly have CMT2 or dHMN.

The contribution of mosaicism to genetic diseases and de novo pathogenic variants.

The contribution of mosaicism to diagnosed genetic disease and presumed de novo variants (DNV) is under investigated. We determined the contribution of mosaic genetic disease (MGD) and diagnosed parental mosaicism (PM) in parents of offspring with reported DNV (in the same variant) in the (1) Undiagnosed Diseases Network (UDN) (N = 1946) and (2) in 12,472 individuals electronic health records (EHR) who underwent genetic testing at an academic medical center. In the UDN, we found 4.51% of diagnosed probands had MGD, and 2.86% of parents of those with DNV exhibited PM. In the EHR, we found 6.03% and 2.99% and (of diagnosed probands) had MGD detected on chromosomal microarray and exome/genome sequencing, respectively. We found 2.34% (of those with a presumed pathogenic DNV) had a parent with PM for the variant. We detected mosaicism (regardless of pathogenicity) in 4.49% of genetic tests performed. We found a broad phenotypic spectrum of MGD with previously unknown phenotypic phenomena. MGD is highly heterogeneous and provides a significant contribution to genetic diseases. Further work is required to improve the diagnosis of MGD and investigate how PM contributes to DNV risk.

Commonalities across computational workflows for uncovering explanatory variants in undiagnosed cases.

PURPOSE: Genomic sequencing has become an increasingly powerful and relevant tool to be leveraged for the discovery of genetic aberrations underlying rare, Mendelian conditions. Although the computational tools incorporated into diagnostic workflows for this task are continually evolving and improving, we nevertheless sought to investigate commonalities across sequencing processing workflows to reveal consensus and standard practice tools and highlight exploratory analyses where technical and theoretical method improvements would be most impactful. METHODS: We collected details regarding the computational approaches used by a genetic testing laboratory and 11 clinical research sites in the United States participating in the Undiagnosed Diseases Network via meetings with bioinformaticians, online survey forms, and analyses of internal protocols. RESULTS: We found that tools for processing genomic sequencing data can be grouped into four distinct categories. Whereas well-established practices exist for initial variant calling and quality control steps, there is substantial divergence across sites in later stages for variant prioritization and multimodal data integration, demonstrating a diversity of approaches for solving the most mysterious undiagnosed cases. CONCLUSION: The largest differences across diagnostic workflows suggest that advances in structural variant detection, noncoding variant interpretation, and integration of additional biomedical data may be especially promising for solving chronically undiagnosed cases.

Clinical sites of the Undiagnosed Diseases Network: unique contributions to genomic medicine and science.

PURPOSE: The NIH Undiagnosed Diseases Network (UDN) evaluates participants with disorders that have defied diagnosis, applying personalized clinical and genomic evaluations and innovative research. The clinical sites of the UDN are essential to advancing the UDN mission; this study assesses their contributions relative to standard clinical practices. METHODS: We analyzed retrospective data from four UDN clinical sites, from July 2015 to September 2019, for diagnoses, new disease gene discoveries and the underlying investigative methods. RESULTS: Of 791 evaluated individuals, 231 received 240 diagnoses and 17 new disease-gene associations were recognized. Straightforward diagnoses on UDN exome and genome sequencing occurred in 35% (84/240). We considered these tractable in standard clinical practice, although genome sequencing is not yet widely available clinically. The majority (156/240, 65%) required additional UDN-driven investigations, including 90 diagnoses that occurred after prior nondiagnostic exome sequencing and 45 diagnoses (19%) that were nongenetic. The UDN-driven investigations included complementary/supplementary phenotyping, innovative analyses of genomic variants, and collaborative science for functional assays and animal modeling. CONCLUSION: Investigations driven by the clinical sites identified diagnostic and research paradigms that surpass standard diagnostic processes. The new diagnoses, disease gene discoveries, and delineation of novel disorders represent a model for genomic medicine and science.

Development and Progression of Radiologic Abnormalities in Individuals at Risk for Familial Interstitial Lung Disease.

Rationale: The preclinical natural history of progressive lung fibrosis is poorly understood.Objectives: Our goals were to identify risk factors for interstitial lung abnormalities (ILA) on high-resolution computed tomography (HRCT) scans and to determine progression toward clinical interstitial lung disease (ILD) among subjects in a longitudinal cohort of self-reported unaffected first-degree relatives of patients with familial interstitial pneumonia.Methods: Enrollment evaluation included a health history and exposure questionnaire and HRCT scans, which were categorized by visual assessment as no ILA, early/mild ILA, or extensive ILA. The study endpoint was met when ILA were extensive or when ILD was diagnosed clinically. Among subjects with adequate study time to complete 5-year follow-up HRCT, the proportion with ILD events (endpoint met or radiographic ILA progression) was calculated.Measurements and Main Results: Among 336 subjects, the mean age was 53.1 (SD, 9.9) years. Those with ILA (early/mild [n = 74] or extensive [n = 3]) were older, were more likely to be ever smokers, had shorter peripheral blood mononuclear cell telomeres, and were more likely to carry the MUC5B risk allele. Self-reported occupational or environmental exposures, including aluminum smelting, lead, birds, and mold, were independently associated with ILA. Among 129 subjects with sufficient study time, 25 (19.4%) had an ILD event by 5 years after enrollment; of these, 12 met the study endpoint and another 13 had radiologic progression of ILA. ILD events were more common among those with early/mild ILA at enrollment (63.3% vs. 6.1%; P < 0.0001).Conclusions: Rare and common environmental exposures are independent risk factors for radiologic abnormalities. In 5 years, progression of ILA occurred in most individuals with early ILA detected at enrollment.

HACER: an atlas of human active enhancers to interpret regulatory variants.

Recent studies have shown that disease-susceptibility variants frequently lie in cell-type-specific enhancer elements. To identify, interpret, and prioritize such risk variants, we must identify the enhancers active in disease-relevant cell types, their upstream transcription factor (TF) binding, and their downstream target genes. To address this need, we built HACER (http://bioinfo.vanderbilt.edu/AE/HACER/), an atlas of Human ACtive Enhancers to interpret Regulatory variants. The HACER atlas catalogues and annotates in-vivo transcribed cell-type-specific enhancers, as well as placing enhancers within transcriptional regulatory networks by integrating ENCODE TF ChIP-Seq and predicted/validated chromatin interaction data. We demonstrate the utility of HACER in (i) offering a mechanistic hypothesis to explain the association of SNP rs614367 with ER-positive breast cancer risk, (ii) exploring tumor-specific enhancers in selective MYC dysregulation and (iii) prioritizing/annotating non-coding regulatory regions targeting CCND1. HACER provides a valuable resource for studies of GWAS, non-coding variants, and enhancer-mediated regulation.