Pathogenicity Of Missense Variants Research Articles

Genetic diversity of 1,845 rhesus macaques improves genetic variation interpretation and identifies disease models

Understanding and treating human diseases require valid animal models. Leveraging the genetic diversity in rhesus macaque populations across eight primate centers in the United States, we conduct targeted-sequencing on 1845 individuals for 374 genes linked to inherited human retinal and neurodevelopmental diseases. We identify over 47,000 single nucleotide variants, a substantial proportion of which are shared with human populations. By combining rhesus and human allele frequencies with established variant prediction methods, we develop a machine learning-based score that outperforms established methods in predicting missense variant pathogenicity. Remarkably, we find a marked number of loss-of-function variants and putative deleterious variants, which may lead to the development of rhesus disease models. Through phenotyping of macaques carrying a pathogenic OPA1:p.A8S variant, we identify a genetic model of autosomal dominant optic atrophy. Finally, we present a public website housing variant and genotype data from over two thousand rhesus macaques.

Clinical utility of AlphaMissense in predicting pathogenicity of DNA damage repair genes in cancer.

10581 Background: Deficiencies in DNA damage repair (DDR) impact cancer predisposition and various treatment responses.Functional implications of missense variants in DDR genes remain elusive.AlphaMissense, a new AI-based method, accurately predicts the pathogenicity of missense variants, but its clinical utility requires validation. We evaluated the accuracy of AlphaMissense predictions by analyzing known mutational signatures, genomic characteristics, and therapeutic vulnerabilities of DDR-deficient tumors. Methods: We analyzed data from TCGA (n=8,178) and MSK-IMPACT targeted panel sequencing (n=56,695) to evaluate the performance of AlphaMissense in predicting clinically actionable DDR genes, including core homologous recombination (HR) genes ( BRCA1/2, PALB2, RAD51C), POLE, and ATM. Missense mutations were classified as known pathogenic by OncoKB or previous literature, newly pathogenic identified with AlphaMisense, or benign. Mutational signatures were computed using SigMA algorithm, classifying each signature as positive at >20%. The irradiated tumor control was evaluated in patients with ATM missense mutations who received radiotherapy. Results: Among 6,743 unique DDR missense mutations from MSK-IMPACT, 372 (3.5%) were known pathogenic, and 1,182 (17.5%) were new pathogenic mutations. In 205 tumors with bi-allelic core-HR missense mutations in breast, ovarian, pancreatic, and prostate cancers, 20 (9.8%) had new pathogenic mutations. Tumors with new pathogenic mutations more frequently exhibited HR-deficient signatures (65%) than wild-type (31.8%, p < .001), similar to tumors with known pathogenic mutations (62.8%, p = 1.0). In TCGA, HR-deficient signatures tended to be more frequent in tumors with new pathogenic missense mutations than in wild-type tumors (41.7% vs. 21.0%, p = .16). Of 1,010 tumors with POLE missense mutations from MSK-IMPACT, 300 (30%) had new pathogenic mutations. The presence of POLE signatures in tumors with new pathogenic mutations (11.9%) was not significantly different from those with benign mutations (4.3%, p = .19) and less frequent than in tumors with known pathogenic mutations (91.0%, p < .001). In MSK-IMPACT, tumors with new pathogenic ATM missense mutations were less likely to have TP53 mutations than benign missense mutations (32.7% vs. 56.6%, p < .001), consistent with bona-fide pathogenicity. Further, patients with new pathogenic ATM missense mutations showed increased radiosensitivity, characteristic of ATM deficiency, and improved irradiated tumor control compared to those with benign missense mutations (Hazard ratio 0.58, 95% Cl 0.35-0.95, p = .03). Conclusions: AlphaMissense can identify previously unknown pathogenic DDR missense mutations, but its accuracy is gene-dependent. AlphaMissense pathogenicity predictions for DDR genes can utilized to help classify mutations, but may not be sufficient on their own.

Analysis of AlphaMissense data in different protein groups and structural context

Single amino acid substitutions can profoundly affect protein folding, dynamics, and function. The ability to discern between benign and pathogenic substitutions is pivotal for therapeutic interventions and research directions. Given the limitations in experimental examination of these variants, AlphaMissense has emerged as a promising predictor of the pathogenicity of missense variants. Since heterogenous performance on different types of proteins can be expected, we assessed the efficacy of AlphaMissense across several protein groups (e.g. soluble, transmembrane, and mitochondrial proteins) and regions (e.g. intramembrane, membrane interacting, and high confidence AlphaFold segments) using ClinVar data for validation. Our comprehensive evaluation showed that AlphaMissense delivers outstanding performance, with MCC scores predominantly between 0.6 and 0.74. We observed low performance on disordered datasets and ClinVar data related to the CFTR ABC protein. However, a superior performance was shown when benchmarked against the high quality CFTR2 database. Our results with CFTR emphasizes AlphaMissense’s potential in pinpointing functional hot spots, with its performance likely surpassing benchmarks calculated from ClinVar and ProteinGym datasets.

Joint genotypic and phenotypic outcome modeling improves base editing variant effect quantification.

CRISPR base editing screens enable analysis of disease-associated variants at scale; however, variable efficiency and precision confounds the assessment of variant-induced phenotypes. Here, we provide an integrated experimental and computational pipeline that improves estimation of variant effects in base editing screens. We use a reporter construct to measure guide RNA (gRNA) editing outcomes alongside their phenotypic consequences and introduce base editor screen analysis with activity normalization (BEAN), a Bayesian network that uses per-guide editing outcomes provided by the reporter and target site chromatin accessibility to estimate variant impacts. BEAN outperforms existing tools in variant effect quantification. We use BEAN to pinpoint common regulatory variants that alter low-density lipoprotein (LDL) uptake, implicating previously unreported genes. Additionally, through saturation base editing of LDLR, we accurately quantify missense variant pathogenicity that is consistent with measurements in UK Biobank patients and identify underlying structural mechanisms. This work provides a widely applicable approach to improve the power of base editing screens for disease-associated variant characterization.

Enhancing missense variant pathogenicity prediction with protein language models using VariPred

Computational approaches for predicting the pathogenicity of genetic variants have advanced in recent years. These methods enable researchers to determine the possible clinical impact of rare and novel variants. Historically these prediction methods used hand-crafted features based on structural, evolutionary, or physiochemical properties of the variant. In this study we propose a novel framework that leverages the power of pre-trained protein language models to predict variant pathogenicity. We show that our approach VariPred (Variant impact Predictor) outperforms current state-of-the-art methods by using an end-to-end model that only requires the protein sequence as input. Using one of the best-performing protein language models (ESM-1b), we establish a robust classifier that requires no calculation of structural features or multiple sequence alignments. We compare the performance of VariPred with other representative models including 3Cnet, Polyphen-2, REVEL, MetaLR, FATHMM and ESM variant. VariPred performs as well as, or in most cases better than these other predictors using six variant impact prediction benchmarks despite requiring only sequence data and no pre-processing of the data.

Functional Studies of Deafness-Associated Pendrin and Prestin Variants.

Pendrin and prestin are evolutionary-conserved membrane proteins that are essential for normal hearing. Dysfunction of these proteins results in hearing loss in humans, and numerous deafness-associated pendrin and prestin variants have been identified in patients. However, the pathogenic impacts of many of these variants are ambiguous. Here, we report results from our ongoing efforts to experimentally characterize pendrin and prestin variants using in vitro functional assays. With previously established fluorometric anion transport assays, we determined that many of the pendrin variants identified on transmembrane (TM) 10, which contains the essential anion binding site, and on the neighboring TM9 within the core domain resulted in impaired anion transport activity. We also determined the range of functional impairment in three deafness-associated prestin variants by measuring nonlinear capacitance (NLC), a proxy for motor function. Using the results from our functional analyses, we also evaluated the performance of AlphaMissense (AM), a computational tool for predicting the pathogenicity of missense variants. AM prediction scores correlated well with our experimental results; however, some variants were misclassified, underscoring the necessity of experimentally assessing the effects of variants. Together, our experimental efforts provide invaluable information regarding the pathogenicity of deafness-associated pendrin and prestin variants.

Benchmarking AlphaMissense pathogenicity predictions against cystic fibrosis variants.

Variants in the cystic fibrosis transmembrane conductance regulator gene (CFTR) result in cystic fibrosis-a lethal autosomal recessive disorder. Missense variants that alter a single amino acid in the CFTR protein are among the most common cystic fibrosis variants, yet tools for accurately predicting molecular consequences of missense variants have been limited to date. AlphaMissense (AM) is a new technology that predicts the pathogenicity of missense variants based on dual learned protein structure and evolutionary features. Here, we evaluated the ability of AM to predict the pathogenicity of CFTR missense variants. AM predicted a high pathogenicity for CFTR residues overall, resulting in a high false positive rate and fair classification performance on CF variants from the CFTR2.org database. AM pathogenicity score correlated modestly with pathogenicity metrics from persons with CF including sweat chloride level, pancreatic insufficiency rate, and Pseudomonas aeruginosa infection rate. Correlation was also modest with CFTR trafficking and folding competency in vitro. By contrast, the AM score correlated well with CFTR channel function in vitro-demonstrating the dual structure and evolutionary training approach learns important functional information despite lacking such data during training. Different performance across metrics indicated AM may determine if polymorphisms in CFTR are recessive CF variants yet cannot differentiate mechanistic effects or the nature of pathophysiology. Finally, AM predictions offered limited utility to inform on the pharmacological response of CF variants i.e., theratype. Development of new approaches to differentiate the biochemical and pharmacological properties of CFTR variants is therefore still needed to refine the targeting of emerging precision CF therapeutics.

Clinical manifestations and spermatogenesis outcomes in Chinese patients with congenital hypogonadotropic hypogonadism caused by inherited or de novo FGFR1 mutations.

Fibroblast growth factor receptor 1 (FGFR1) mutations are associated with congenital hypogonadotropic hypogonadism (CHH) through inheritance or spontaneous occurrence. We detected FGFR1 mutations in a Chinese cohort of 210 CHH patients at Peking Union Medical College Hospital (Beijing, China) using next-generation and Sanger sequencing. We assessed missense variant pathogenicity using six bioinformatics tools and compared clinical features and treatment outcomes between inherited and de novo mutation groups. Among 19 patients with FGFR1 mutations, three were recurrent, and 16 were novel variants. Sixteen of the novel mutations were likely pathogenic according to the American College of Medical Genetics and Genomics (ACMG) guidelines, with the prevalent P366L variant. The majority of FGFR1 mutations was inherited (57.9%), with frameshift mutations exclusive to the de novo mutation group. The inherited mutation group had a lower incidence of cryptorchidism, short stature, and skeletal deformities. In the inherited mutation group, luteinizing hormone (LH) levels were 0.5 IU l-1, follicle-stimulating hormone (FSH) levels were 1.0 IU l-1, and testosterone levels were 1.3 nmol l-1. In contrast, the de novo group had LH levels of 0.2 IU l-1, FSH levels of 0.5 IU l-1, and testosterone levels of 0.9 nmol l-1, indicating milder hypothalamus-pituitary-gonadal axis (HPGA) functional deficiency in the inherited group. The inherited mutation group showed a tendency toward higher spermatogenesis rates. In conclusion, this study underscores the predominance of inherited FGFR1 mutations and their association with milder HPGA dysfunction compared to de novo mutations, contributing to our understanding of the genetic and clinical aspects of FGFR1 mutations.

SIGMA leverages protein structural information to predict the pathogenicity of missense variants

Leveraging protein structural information to evaluate pathogenicity has been hindered by the scarcity of experimentally determined 3D protein. With the aid of AlphaFold2 predictions, we developed the structure-informed genetic missense mutation assessor (SIGMA) to predict missense variant pathogenicity. In comparison with existing predictors across labeled variant datasets and experimental datasets, SIGMA demonstrates superior performance in predicting missense variant pathogenicity (AUC= 0.933). We found that the relative solvent accessibility of the mutated residue contributed greatly to the predictive ability of SIGMA. We further explored combining SIGMA with other top-tier predictors to create SIGMA+, proving highly effective for variant pathogenicity prediction (AUC= 0.966). To facilitate the application of SIGMA, we pre-computed SIGMA scores for over 48 million possible missense variants across 3,454 disease-associated genes and developed an interactive online platform (https://www.sigma-pred.org/). Overall, by leveraging protein structure information, SIGMA offers an accurate structure-based approach to evaluating the pathogenicity of missense variants.

REVEL Is Better at Predicting Pathogenicity of Loss-of-Function than Gain-of-Function Variants

In silico predictive tools can help determine the pathogenicity of variants. The 2015 American College of Medical Genetics and Genomics (ACMG) guidelines recommended that scores from these tools can be used as supporting evidence of pathogenicity. A subsequent publication by the ClinGen Sequence Variant Interpretation Working Group suggested that high scores from some tools were sufficiently predictive to be used as moderate or strong evidence of pathogenicity. REVEL is a widely used metapredictor that uses the scores of 13 individual in silico tools to calculate the pathogenicity of missense variants. Its ability to predict missense pathogenicity has been assessed extensively; however, no study has previously tested whether its performance is affected by whether the missense variant acts via a loss-of-function (LoF) or gain-of-function (GoF) mechanism. We used a highly curated dataset of 66 confirmed LoF and 65 confirmed GoF variants to evaluate whether this affected the performance of REVEL. 98% of LoF and 100% of GoF variants met the author-recommended REVEL threshold of 0.5 for pathogenicity, while 89% of LoF and 88% of GoF variants exceeded the 0.75 threshold. However, while 55% of LoF variants met the threshold recommended for a REVEL score to count as strong evidence of pathogenicity from the ACMG guidelines (0.932), only 35% of GoF variants met this threshold ( P = 0.0352 ). GoF variants are therefore less likely to receive the highest REVEL scores which would enable the REVEL score to be used as strong evidence of pathogenicity. This has implications for classification with the ACMG guidelines as GoF variants are less likely to meet the criteria for pathogenicity.

MmisAT and MmisP: an efficient and accurate suite of variant analysis toolkit for primary mitochondrial diseases

Recent advances in next-generation sequencing (NGS) technology have greatly accelerated the need for efficient annotation to accurately interpret clinically relevant genetic variants in human diseases. Therefore, it is crucial to develop appropriate analytical tools to improve the interpretation of disease variants. Given the unique genetic characteristics of mitochondria, including haplogroup, heteroplasmy, and maternal inheritance, we developed a suite of variant analysis toolkits specifically designed for primary mitochondrial diseases: the Mitochondrial Missense Variant Annotation Tool (MmisAT) and the Mitochondrial Missense Variant Pathogenicity Predictor (MmisP). MmisAT can handle protein-coding variants from both nuclear DNA and mtDNA and generate 349 annotation types across six categories. It processes 4.78 million variant data in 76 min, making it a valuable resource for clinical and research applications. Additionally, MmisP provides pathogenicity scores to predict the pathogenicity of genetic variations in mitochondrial disease. It has been validated using cross-validation and external datasets and demonstrated higher overall discriminant accuracy with a receiver operating characteristic (ROC) curve area under the curve (AUC) of 0.94, outperforming existing pathogenicity predictors. In conclusion, the MmisAT is an efficient tool that greatly facilitates the process of variant annotation, expanding the scope of variant annotation information. Furthermore, the development of MmisP provides valuable insights into the creation of disease-specific, phenotype-specific, and even gene-specific predictors of pathogenicity, further advancing our understanding of specific fields.

Predicting the pathogenicity of missense variants based on protein instability to support diagnosis of patients with novel variants of ARSL

Rare diseases are estimated to affect 3.5%–5.9% of the population worldwide and are difficult to diagnose. Genome analysis is useful for diagnosis. However, since some variants, especially missense variants, are also difficult to interpret, tools to accurately predict the effect of missense variants are very important and needed. Here we developed a method, “VarMeter”, to predict whether a missense variant is damaging based on Gibbs free energy and solvent-accessible surface area calculated from the AlphaFold 3D protein model. We applied this method to the whole-exome sequencing data of 900 individuals with rare or undiagnosed disease in our in-house database, and identified four who were hemizygous for missense variants of arylsulfatase L (ARSL; known as the genetic cause of chondrodysplasia punctata 1, CPDX1). Two individuals had a novel Ser89 to Asn (Ser89Asn) or Arg469 to Trp (Arg469Trp) substitution, respectively predicted as “damaging” or “benign”; the other two had an Arg111 to His (Arg111His) or Gly117 to Arg (Gly117Arg) substitution, respectively predicted as “damaging” or “possibly damaging” and previously reported in patients showing clinical manifestations of CDPX1. Expression and analysis of the missense variant proteins showed that the predicted pathogenic variants (Ser89Asn, Arg111His, and Gly117Arg) had complete loss of sulfatase activity and reduced protease resistance due to destabilization of protein structure, while the predicted benign variant (Arg469Trp) had activity and protease resistance comparable to those of wild-type ARSL. The individual with the novel pathogenic Ser89Asn variant exhibited characteristics of CDPX1, while the individual with the benign Arg469Trp variant exhibited no such characteristics. These findings demonstrate that VarMeter may be used to predict the deleteriousness of variants found in genome sequencing data and thereby support disease diagnosis.

Predictive Modeling and Structure Analysis of Genetic Variants in Familial Hypercholesterolemia: Implications for Diagnosis and Protein Interaction Studies.

Familial hypercholesterolemia (FH) is a hereditary condition characterized by elevated levels of low-density lipoprotein cholesterol (LDL-C), which increases the risk of cardiovascular disease if left untreated. This review aims to discuss the role of bioinformatics tools in evaluating the pathogenicity of missense variants associated with FH. Specifically, it highlights the use of predictive models based on protein sequence, structure, evolutionary conservation, and other relevant features in identifying genetic variants within LDLR, APOB, and PCSK9 genes that contribute to FH. In recent years, various bioinformatics tools have emerged as valuable resources for analyzing missense variants in FH-related genes. Tools such as REVEL, Varity, and CADD use diverse computational approaches to predict the impact of genetic variants on protein function. These tools consider factors such as sequence conservation, structural alterations, and receptor binding to aid in interpreting the pathogenicity of identified missense variants. While these predictive models offer valuable insights, the accuracy of predictions can vary, especially for proteins with unique characteristics that might not be well represented in the databases used for training. This review emphasizes the significance of utilizing bioinformatics tools for assessing the pathogenicity of FH-associated missense variants. Despite their contributions, a definitive diagnosis of a genetic variant necessitates functional validation through in vitro characterization or cascade screening. This step ensures the precise identification of FH-related variants, leading to more accurate diagnoses. Integrating genetic data with reliable bioinformatics predictions and functional validation can enhance our understanding of the genetic basis of FH, enabling improved diagnosis, risk stratification, and personalized treatment for affected individuals. The comprehensive approach outlined in this review promises to advance the management of this inherited disorder, potentially leading to better health outcomes for those affected by FH.

Advancing missense variant pathogenicity prediction.

Advancing missense variant pathogenicity prediction.

Accurate proteome-wide missense variant effect prediction with AlphaMissense.

The vast majority of missense variants observed in the human genome are of unknown clinical significance. We present AlphaMissense, an adaptation of AlphaFold fine-tuned on human and primate variant population frequency databases to predict missense variant pathogenicity. By combining structural context and evolutionary conservation, our model achieves state-of-the-art results across a wide range of genetic and experimental benchmarks, all without explicitly training on such data. The average pathogenicity score of genes is also predictive for their cell essentiality, capable of identifying short essential genes that existing statistical approaches are underpowered to detect. As a resource to the community, we provide a database of predictions for all possible human single amino acid substitutions and classify 89% of missense variants as either likely benign or likely pathogenic.

Electroclinical Features in Two Novel STRADA Patients and a Functional Yeast Assay for the Validation of Missense STRADA Mutations

Loss of function of the STRADA gene, an upstream mTOR inhibitor, causes a rare neurodevelopmental disorder characterized by polyhydramnios, megalencephaly, and symptomatic epilepsy (PMSE syndrome). Patients display a homogeneous phenotype including early-onset drug-resistant epilepsy, severe psychomotor delay, multisystemic comorbidities, and increased risk of premature death. The administration of sirolimus, an mTOR inhibitor, is helpful in controlling seizures in this syndrome. We report the electroclinical phenotype of two novel patients and the development of a yeast model to validate the pathogenicity of missense variants. Patient 1 harbored a missense STRADA variant and had a peculiar electroclinical phenotype with a relatively mild epilepsy course. Patient 2 harbored a truncating STRADA variant and showed a typical PMSE phenotype and a favorable response to early treatment with sirolimus.When we modeled the p.(Ser264Arg) STRADA change in its yeast homolog SPS1, it impaired SPS1 function.The results underlie the importance of a timely molecular diagnosis in these patients and show that yeast is a simple yet effective model to validate the pathogenicity of missense variants.

Understanding structure-guided variant effect predictions using 3D convolutional neural networks.

Predicting pathogenicity of missense variants in molecular diagnostics remains a challenge despite the available wealth of data, such as evolutionary information, and the wealth of tools to integrate that data. We describe DeepRank-Mut, a configurable framework designed to extract and learn from physicochemically relevant features of amino acids surrounding missense variants in 3D space. For each variant, various atomic and residue-level features are extracted from its structural environment, including sequence conservation scores of the surrounding amino acids, and stored in multi-channel 3D voxel grids which are then used to train a 3D convolutional neural network (3D-CNN). The resultant model gives a probabilistic estimate of whether a given input variant is disease-causing or benign. We find that the performance of our 3D-CNN model, on independent test datasets, is comparable to other widely used resources which also combine sequence and structural features. Based on the 10-fold cross-validation experiments, we achieve an average accuracy of 0.77 on the independent test datasets. We discuss the contribution of the variant neighborhood in the model's predictive power, in addition to the impact of individual features on the model's performance. Two key features: evolutionary information of residues in the variant neighborhood and their solvent accessibilities were observed to influence the predictions. We also highlight how predictions are impacted by the underlying disease mechanisms of missense mutations and offer insights into understanding these to improve pathogenicity predictions. Our study presents aspects to take into consideration when adopting deep learning approaches for protein structure-guided pathogenicity predictions.

Characterizing proteomic and transcriptomic features of missense variants in amyotrophic lateral sclerosis genes.

Within recent years, there has been a growing number of genes associated with amyotrophic lateral sclerosis (ALS), resulting in an increasing number of novel variants, particularly missense variants, many of which are of unknown clinical significance. Here, we leverage the sequencing efforts of the ALS Knowledge Portal (3,864 individuals with ALS and 7,839 controls) and Project MinE ALS Sequencing Consortium (4,366 individuals with ALS and 1,832 controls) to perform proteomic and transcriptomic characterization of missense variants in 24 ALS-associated genes. The two sequencing datasets were interrogated for missense variants in the 24 genes, and variants were annotated with genomic database minor allele frequencies, ClinVar pathogenicity classifications, protein sequence features including Uniprot functional site annotations and PhosphoSitePlus post-translational modification (PTM) site annotations, structural features from AlphaFold predicted monomeric 3D structures, and transcriptomic expression levels from Genotype-Tissue Expression (GTEx). We then applied missense variant enrichment and gene-burden testing following binning of variation based on the selected proteomic and transcriptomic features to identify those most relevant to pathogenicity in ALS-associated genes. Using predicted human protein structures from AlphaFold, we determined that missense variants carried by individuals with ALS were significantly enriched in β-sheets and α-helices, as well as in core, buried, or moderately buried regions. At the same time, we identified that hydrophobic amino acid residues, compositionally biased protein regions and protein-protein interaction regions are predominantly enriched in missense variants carried by individuals with ALS. Assessment of expression level based on transcriptomics also revealed enrichment of variants of high and medium expression across all tissues and within the brain. We further explored enriched features of interest using burden analyses and identified individual genes were indeed driving certain enrichment signals. A case study is presented for SOD1 to demonstrate proof of concept of how enriched features may aid in defining variant pathogenicity. Our results present proteomic and transcriptomic features that are important indicators of missense variant pathogenicity in ALS and are distinct from features associated with neurodevelopmental disorders.

Comprehensive functional characterization of SGCB coding variants predicts pathogenicity in limb-girdle muscular dystrophy type R4/2E.

Genetic testing is essential for patients with a suspected hereditary myopathy. More than 50% of patients clinically diagnosed with a myopathy carry a variant of unknown significance in a myopathy gene, often leaving them without a genetic diagnosis. Limb-girdle muscular dystrophy (LGMD) type R4/2E is caused by mutations in β-sarcoglycan (SGCB). Together, β-, α-, γ-, and δ-sarcoglycan form a 4-protein transmembrane complex (SGC) that localizes to the sarcolemma. Biallelic loss-of-function mutations in any subunit can lead to LGMD. To provide functional evidence for the pathogenicity of missense variants, we performed deep mutational scanning of SGCB and assessed SGC cell surface localization for all 6,340 possible amino acid changes. Variant functional scores were bimodally distributed and perfectly predicted pathogenicity of known variants. Variants with less severe functional scores more often appeared in patients with slower disease progression, implying a relationship between variant function and disease severity. Amino acid positions intolerant to variation mapped to points of predicted SGC interactions, validated in silico structural models, and enabled accurate prediction of pathogenic variants in other SGC genes. These results will be useful for clinical interpretation of SGCB variants and improving diagnosis of LGMD; we hope they enable wider use of potentially life-saving gene therapy.

Accuracy of a machine learning method based on structural and locational information from AlphaFold2 for predicting the pathogenicity of TARDBP and FUS gene variants in ALS

BackgroundIn the sporadic form of amyotrophic lateral sclerosis (ALS), the pathogenicity of rare variants in the causative genes characterizing the familial form remains largely unknown. To predict the pathogenicity of such variants, in silico analysis is commonly used. In some ALS causative genes, the pathogenic variants are concentrated in specific regions, and the resulting alterations in protein structure are thought to significantly affect pathogenicity. However, existing methods have not taken this issue into account. To address this, we have developed a technique termed MOVA (method for evaluating the pathogenicity of missense variants using AlphaFold2), which applies positional information for structural variants predicted by AlphaFold2. Here we examined the utility of MOVA for analysis of several causative genes of ALS.MethodsWe analyzed variants of 12 ALS-related genes (TARDBP, FUS, SETX, TBK1, OPTN, SOD1, VCP, SQSTM1, ANG, UBQLN2, DCTN1, and CCNF) and classified them as pathogenic or neutral. For each gene, the features of the variants, consisting of their positions in the 3D structure predicted by AlphaFold2, pLDDT score, and BLOSUM62 were trained into a random forest and evaluated by the stratified fivefold cross validation method. We compared how accurately MOVA predicted mutant pathogenicity with other in silico prediction methods and evaluated the prediction accuracy at TARDBP and FUS hotspots. We also examined which of the MOVA features had the greatest impact on pathogenicity discrimination.ResultsMOVA yielded useful results (AUC ≥ 0.70) for TARDBP, FUS, SOD1, VCP, and UBQLN2 of 12 ALS causative genes. In addition, when comparing the prediction accuracy with other in silico prediction methods, MOVA obtained the best results among those compared for TARDBP, VCP, UBQLN2, and CCNF. MOVA demonstrated superior predictive accuracy for the pathogenicity of mutations at hotspots of TARDBP and FUS. Moreover, higher accuracy was achieved by combining MOVA with REVEL or CADD. Among the features of MOVA, the x, y, and z coordinates performed the best and were highly correlated with MOVA.ConclusionsMOVA is useful for predicting the virulence of rare variants in which they are concentrated at specific structural sites, and for use in combination with other prediction methods.