Explicable prioritization of genetic variants by integration of rule-based and machine learning algorithms for diagnosis of rare Mendelian disorders

BACKGROUND

In the process of finding the causative variant of rare diseases, accurate assessment and prioritization of genetic variants is essential. Previous variant prioritization tools mainly depend on the in-silico prediction of the pathogenicity of variants, which results in low sensitivity and difficulty in interpreting the prioritization result. In this study, we propose an explainable algorithm for variant prioritization, named 3ASC, with higher sensitivity and ability to annotate evidence used for prioritization. 3ASC annotates each variant with the 28 criteria defined by the ACMG/AMP genome interpretation guidelines and features related to the clinical interpretation of the variants. The system can explain the result based on annotated evidence and feature contributions.

RESULTS

We trained various machine learning algorithms using in-house patient data. The performance of variant ranking was assessed using the recall rate of identifying causative variants in the top-ranked variants. The best practice model was a random forest classifier that showed top 1 recall of 85.6% and top 3 recall of 94.4%. The 3ASC annotates the ACMG/AMP criteria for each genetic variant of a patient so that clinical geneticists can interpret the result as in the CAGI6 SickKids challenge. In the challenge, 3ASC identified causal genes for 10 out of 14 patient cases, with evidence of decreased gene expression for 6 cases. Among them, two genes (HDAC8 and CASK) had decreased gene expression profiles confirmed by transcriptome data.

CONCLUSIONS

3ASC can prioritize genetic variants with higher sensitivity compared to previous methods by integrating various features related to clinical interpretation, including features related to false positive risk such as quality control and disease inheritance pattern. The system allows interpretation of each variant based on the ACMG/AMP criteria and feature contribution assessed using explainable AI techniques.

GPAD: a natural language processing-based application to extract the gene-disease association discovery information from OMIM

BACKGROUND

Thousands of genes have been associated with different Mendelian conditions. One of the valuable sources to track these gene-disease associations (GDAs) is the Online Mendelian Inheritance in Man (OMIM) database. However, most of the information in OMIM is textual, and heterogeneous (e.g. summarized by different experts), which complicates automated reading and understanding of the data. Here, we used Natural Language Processing (NLP) to make a tool (Gene-Phenotype Association Discovery (GPAD)) that could syntactically process OMIM text and extract the data of interest.

RESULTS

GPAD applies a series of language-based techniques to the text obtained from OMIM API to extract GDA discovery-related information. GPAD can inform when a particular gene was associated with a specific phenotype, as well as the type of validation-whether through model organisms or cohort-based patient-matching approaches-for such an association. GPAD extracted data was validated with published reports and was compared with large language model. Utilizing GPAD's extracted data, we analysed trends in GDA discoveries, noting a significant increase in their rate after the introduction of exome sequencing, rising from an average of about 150-250 discoveries each year. Contrary to hopes of resolving most GDAs for Mendelian disorders by now, our data indicate a substantial decline in discovery rates over the past five years (2017-2022). This decline appears to be linked to the increasing necessity for larger cohorts to substantiate GDAs. The rising use of zebrafish and Drosophila as model organisms in providing evidential support for GDAs is also observed.

CONCLUSIONS

GPAD's real-time analyzing capacity offers an up-to-date view of GDA discovery and could help in planning and managing the research strategies. In future, this solution can be extended or modified to capture other information in OMIM and scientific literature.

Variants in ACTC1 underlie distal arthrogryposis accompanied by congenital heart defects

Contraction of the human sarcomere is the result of interactions between myosin cross-bridges and actin filaments. Pathogenic variants in genes such as MYH7, TPM1, and TNNI3 that encode parts of the cardiac sarcomere cause muscle diseases that affect the heart, such as dilated cardiomyopathy and hypertrophic cardiomyopathy. In contrast, pathogenic variants in homologous genes such as MYH2, TPM2, and TNNI2 that encode parts of the skeletal muscle sarcomere cause muscle diseases affecting skeletal muscle, such as distal arthrogryposis (DA) syndromes and skeletal myopathies. To date, there have been few reports of genes (e.g., MYH7) encoding sarcomeric proteins in which the same pathogenic variant affects skeletal and cardiac muscle. Moreover, none of the known genes underlying DA have been found to contain pathogenic variants that also cause cardiac abnormalities. We report five families with DA because of heterozygous missense variants in the gene actin, alpha, cardiac muscle 1 (ACTC1). ACTC1 encodes a highly conserved actin that binds to myosin in cardiac and skeletal muscle. Pathogenic variants in ACTC1 have been found previously to underlie atrial septal defect, dilated cardiomyopathy, hypertrophic cardiomyopathy, and left ventricular noncompaction. Our discovery delineates a new DA condition because of variants in ACTC1 and suggests that some functions of ACTC1 are shared in cardiac and skeletal muscle.

Bi-allelic SNAPC4 variants dysregulate global alternative splicing and lead to neuroregression and progressive spastic paraparesis

The vast majority of human genes encode multiple isoforms through alternative splicing, and the temporal and spatial regulation of those isoforms is critical for organismal development and function. The spliceosome, which regulates and executes splicing reactions, is primarily composed of small nuclear ribonucleoproteins (snRNPs) that consist of small nuclear RNAs (snRNAs) and protein subunits. snRNA gene transcription is initiated by the snRNA-activating protein complex (SNAPc). Here, we report ten individuals, from eight families, with bi-allelic, deleterious SNAPC4 variants. SNAPC4 encoded one of the five SNAPc subunits that is critical for DNA binding. Most affected individuals presented with delayed motor development and developmental regression after the first year of life, followed by progressive spasticity that led to gait alterations, paraparesis, and oromotor dysfunction. Most individuals had cerebral, cerebellar, or basal ganglia volume loss by brain MRI. In the available cells from affected individuals, SNAPC4 abundance was decreased compared to unaffected controls, suggesting that the bi-allelic variants affect SNAPC4 accumulation. The depletion of SNAPC4 levels in HeLa cell lines via genomic editing led to decreased snRNA expression and global dysregulation of alternative splicing. Analysis of available fibroblasts from affected individuals showed decreased snRNA expression and global dysregulation of alternative splicing compared to unaffected cells. Altogether, these data suggest that these bi-allelic SNAPC4 variants result in loss of function and underlie the neuroregression and progressive spasticity in these affected individuals.

Using deep learning and electronic health records to detect Noonan syndrome in pediatric patients

PURPOSE

The variable expressivity and multisystem features of Noonan syndrome (NS) make it difficult for patients to obtain a timely diagnosis. Genetic testing can confirm a diagnosis, but underdiagnosis is prevalent owing to a lack of recognition and referral for testing. Our study investigated the utility of using electronic health records (EHRs) to identify patients at high risk of NS.

METHODS

Using diagnosis texts extracted from Cincinnati Children's Hospital's EHR database, we constructed deep learning models from 162 NS cases and 16,200 putative controls. Performance was evaluated on 2 independent test sets, one containing patients with NS who were previously diagnosed and the other containing patients with undiagnosed NS.

RESULTS

Our novel method performed significantly better than the previous method, with the convolutional neural network model achieving the highest area under the precision-recall curve in both test sets (diagnosed: 0.43, undiagnosed: 0.16).

CONCLUSION

The results suggested the validity of using text-based deep learning methods to analyze EHR and showed the value of this approach as a potential tool to identify patients with features of rare diseases. Given the paucity of medical geneticists, this has the potential to reduce disease underdiagnosis by prioritizing patients who will benefit most from a genetics referral.

Clinical Exome Sequencing of 1000 Families with Complex Immune Phenotypes: Towards comprehensive genomic evaluations

BACKGROUND

Prospective genetic evaluation of patients at our referral research hospital presents clinical research challenges.

OBJECTIVE

This study sought not only a single-gene explanation for participants' immune-related presentations, but viewed each participant holistically, with the potential to have multiple genetic contributions to their immune-phenotype and other heritable comorbidities relevant to their presentation and health.

METHODS

We developed a program integrating exome sequencing, chromosomal microarray, phenotyping, results return with genetic counseling, and reanalysis in 1505 individuals from 1000 families with suspected or known inborn errors of immunity.

RESULTS

Probands were 50.8% female, 71.5% >18 years, and had diverse immune presentations. Overall, 327/1000 probands (32.7%) received 361 molecular diagnoses. These included 17 probands with diagnostic copy number variants, 32 probands with secondary findings, and 31 probands with multiple molecular diagnoses. Reanalysis added 22 molecular diagnoses, predominantly due to new disease-gene associations (9/22, 40.9%). One-quarter of the molecular diagnoses (92/361) did not involve immune-associated genes. Molecular diagnosis was correlated with younger age, male sex, and a higher number of organ systems involved. This program also facilitated the discovery of new gene-disease associations such as SASH3-related immunodeficiency. A review of treatment options and ClinGen actionability curations suggest that at least 251/361 (69.5%) of these molecular diagnoses could translate into >1 management option.

CONCLUSION

This program contributes to our understanding of the diagnostic and clinical utility whole exome analysis on a large scale.

CLINICAL

Implication: Comprehensive analysis of exome data has diagnostic and clinical utility for patients with suspected inborn errors of immunity.

Challenges and Considerations in Sequence Variant Interpretation for Mendelian Disorders

In 2015, the American College of Medical Genetics and Genomics (ACMG), together with the Association for Molecular Pathology (AMP), published the latest guidelines for the interpretation of sequence variants, which have been widely adopted into clinical practice. Despite these standardized efforts, the degrees of subjectivity and uncertainty allowed by the guidelines can lead to inconsistent variant classification across clinical laboratories, making it difficult to assess the pathogenicity of identified variants. We describe the critical elements of variant interpretation processes and potential pitfalls through practical examples and provide updated information based on a review of recent literature. The variant classification we describe is meant to be applicable to sequence variants for Mendelian disorders, whether identified by single-gene tests, multi-gene panels, exome sequencing, or genome sequencing. Continuing efforts to improve the reproducibility and objectivity of sequence variant interpretation across individuals and laboratories are needed.

Estimating yields of prenatal carrier screening and implications for design of expanded carrier screening panels

PURPOSE

Prenatal genetic carrier screening can identify parents at risk of having a child affected by a recessive condition. However, the conditions/genes most appropriate for screening remain a matter of debate. Estimates of carrier rates across genes are needed to guide construction of carrier screening panels.

METHOD

We leveraged an exome sequencing database (n = 123,136) to estimate carrier rates across six major ancestries for 415 genes associated with severe recessive conditions.

RESULTS

We found that 32.6% (East Asian) to 62.9% (Ashkenazi Jewish) of individuals are variant carriers in at least one of the 415 genes. For couples, screening all 415 genes would identify 0.17-2.52% of couples as being at risk for having a child affected by one of these conditions. Screening just the 40 genes with carrier rate >1.0% would identify more than 76% of these at-risk couples. An ancestry-specific panel designed to capture genes with carrier rates >1.0% would include 5 to 28 genes, while a comparable panethnic panel would include 40 genes.

CONCLUSION

Our work guides the design of carrier screening panels and provides data to assist in counseling prospective parents. Our results highlight a high cumulative carrier rate across genes, underscoring the need for careful selection of genes for screening.

A homozygous loss-of-function CAMK2A mutation causes growth delay, frequent seizures and severe intellectual disability

Calcium/calmodulin-dependent protein kinase II (CAMK2) plays fundamental roles in synaptic plasticity that underlies learning and memory. Here, we describe a new recessive neurodevelopmental syndrome with global developmental delay, seizures and intellectual disability. Using linkage analysis and exome sequencing, we found that this disease maps to chromosome 5q31.1-q34 and is caused by a biallelic germline mutation in CAMK2A. The missense mutation, p.His477Tyr is located in the CAMK2A association domain that is critical for its function and localization. Biochemically, the p.His477Tyr mutant is defective in self-oligomerization and unable to assemble into the multimeric holoenzyme.In vivo, CAMK2AH477Y failed to rescue neuronal defects in C. elegans lacking unc-43, the ortholog of human CAMK2A. In vitro, neurons derived from patient iPSCs displayed profound synaptic defects. Together, our data demonstrate that a recessive germline mutation in CAMK2A leads to neurodevelopmental defects in humans and suggest that dysfunctional CAMK2 paralogs may contribute to other neurological disorders.

Biallelic WRN Mutations in Newly Identified Japanese Werner Syndrome Patients

Werner syndrome (WS) is a rare autosomal recessive disorder characterized by systemic accelerated aging. It is caused by pathogenic variants of the WRN gene that encodes a nuclear helicase. In this report, we describe 4 newly identified WS cases among those referred to the Japanese Werner Consortium, Chiba University, Japan. All 4 cases were compound heterozygotes of the Japanese founder mutation, c.3139-1G>C, and a novel null pathogenic variant, c.1587G>A, c.2448+1G>A, or c.3233+1G>T, or an amino acid substitution variant, c.1720G>A, p.Gly574Arg. These 3 null pathogenic variants were not previously described. The p. Gly574Arg was previously reported in a European patient, and the identification of the second p. Gly574Arg case, with classical WS features, further confirmed the pathogenic nature of this variant. For the case with c.3233+1G>T, we determined the phase of 2 disease-causing mutations and demonstrated that they are on different chromosomes. This assay would be particularly important for those cases with ambiguous clinical diagnosis.

FAM65B is a membrane-associated protein of hair cell stereocilia required for hearing

In a large consanguineous Turkish kindred with recessive nonsyndromic, prelingual, profound hearing loss, we identified in the gene FAM65B (MIM611410) a splice site mutation (c.102-1G>A) that perfectly cosegregates with the phenotype in the family. The mutation leads to exon skipping and deletion of 52-amino acid residues of a PX membrane localization domain. FAM65B is known to be involved in myotube formation and in regulation of cell adhesion, polarization, and migration. We show that wild-type Fam65b is expressed during embryonic and postnatal development stages in murine cochlea, and that the protein localizes to the plasma membranes of the stereocilia of inner and outer hair cells of the inner ear. The wild-type protein targets the plasma membrane, whereas the mutant protein accumulates in cytoplasmic inclusion bodies and does not reach the membrane. In zebrafish, knockdown of fam65b leads to significant reduction of numbers of saccular hair cells and neuromasts and to hearing loss. We conclude that FAM65B is a plasma membrane-associated protein of hair cell stereocilia that is essential for hearing.