Objective: Sudden sensorineural hearing loss (SSHL) is a multifactorial disorder with high heterogeneity, thus the outcomes vary widely. This study aimed to develop predictive models based on four machine learning methods for SSHL, identifying the best performer for clinical application.Design: Single‐centre retrospective study.Setting: Chinese People's liberation army (PLA) hospital, Beijing, China.Participants: A total of 1220 in‐patient SSHL patients were enrolled between June 2008 and December 2015.Main outcome measures: An advanced deep learning technique, deep belief network (DBN), together with the conventional logistic regression (LR), support vector machine (SVM) and multilayer perceptron (MLP) were developed to predict the dichotomised hearing outcome of SSHL by inputting six feature collections derived from 149 potential predictors. Accuracy, precision, recall, F‐score and the area under the receiver operator characteristic curves (ROC‐AUC) were exploited to compare the prediction performance of different models.Results: Overall the best predictive ability was provided by the DBN model when tested in the raw data set with 149 variables, achieving an accuracy of 77.58% and AUC of 0.84. Nevertheless, DBN yielded inferior performance after feature pruning. In contrast, the LR, SVM and MLP models demonstrated opposite trend as the greatest individual prediction powers were obtained when included merely three variables, with the ROC‐AUC ranging from 0.79 to 0.81, and then decreased with the increasing size of input features combinations.Conclusions: With the input of enough features, DBN can be a robust prediction tool for SSHL. But LR is more practical for early prediction in routine clinical application using three readily available variables, that is time elapse between symptom onset and study entry, initial hearing level and audiogram.

OTOF mutations are the principal causes of auditory neuropathy. There are reports on Otof‐related gene therapy in mice, but there is no preclinical research on the drug evaluations. Here, Anc80L65 and the mouse hair cell‐specific Myo15 promoter (mMyo15) are used to selectively and effectively deliver human OTOF to hair cells in mice and nonhuman primates to evaluate the efficacy and safety of OTOF gene therapy drugs. A new dual‐AAV‐OTOF‐hybrid strategy to transfer full‐length OTOF is generated, which can stably restore hearing in adult OTOF^{p.Q939*/Q939*} mice with profound deafness, with the longest duration being at least 150 days, and the best therapeutic effect without difference in hearing from wild‐type mice. An AAV microinjection method into the cochlea of cynomolgus monkeys without hearing impairment is further established and found the OTOF can be safely and effectively driven by the mMyo15 promoter in hair cells. In addition, the therapeutic dose of AAV drugs has no impact on normal hearing and does not cause significant systemic toxicity both in mouse and nonhuman primates. In summary, this study develops a potential gene therapy strategy for DFNB9 patients in the clinic and provides complete, standardized, and systematic research data for clinical research and application.

Mutations in the transmembrane channel-like gene 1 (TMC1) can cause both DFNA36 and DFNB7/11 hearing loss. More than thirty DFNB7/11 mutations have been reported, but only three DFNA36 mutations were reported previously. In this study, we found a large Chinese family with 222 family members showing post-lingual, progressive sensorineural hearing loss which were consistent with DFNA36 hearing loss. Auditory brainstem response (ABR) test of the youngest patient showed a special result with nearly normal threshold but prolonged latency, decreased amplitude, and the abnormal waveform morphology. Exome sequencing of the proband found four candidate variants in known hearing loss genes. Sanger sequencing in all family members found a novel variant c.1253T>A (p.M418K) in TMC1 at DFNA36 that co-segregated with the phenotype. This mutation in TMC1 is orthologous to the mutation found in the hearing loss mouse model named Bth ten years ago. In another 51 Chinese autosomal dominant hearing loss families, we screened the segments containing the dominant mutations of TMC1 and no functional variants were found. TMC1 is expressed in the hair cells in inner ear. Given the already known roles of TMC1 in the mechanotransduction in the cochlea and its expression in inner ear, our results may provide an interesting perspective into its function in inner ear.

Mutations in OTOF gene, encoding otoferlin, cause DFNB9 deafness and non-syndromic auditory neuropathy (AN). The aim of this study is to identify OTOF mutations in Chinese patients with non-syndromic auditory neuropathy. 73 unrelated Chinese Han patients with AN, including one case of temperature sensitive non-syndromic auditory neuropathy (TS-NSRAN) and 92 ethnicity-matched controls with normal hearing were screened. Forty-five pairs of PCR primers were designed to amplify all of the exons and their flanking regions of the OTOF gene. The PCR products were sequenced and analyzed for mutation identification. Five novel possibly pathogenic variants (c.1740delC, c.2975_2978delAG, c.1194T>A, c.1780G>A, c.4819C > T) were identified in the group of 73 AN patients, in which two novel mutant alleles (c.2975_2978delAG + c.4819C > T) were identified in one Chinese TS-NSRAN case. Besides, 10 non-pathogenic variants of the OTOF gene were found in AN patients and controls. Screening revealed that mutations in the OTOF gene account for AN in 4 of 73(5.5%) sporadic AN patients, which shows a lower genetic load of that gene in contrast to the previous studies based on other populations. Notably, we found two novel mutant alleles related to temperature sensitive non-syndromic auditory neuropathy. This mutation screening study further confirms that the OTOF gene contributes to ANs and to TS-NSRAN.

In this study, a five-generation Chinese family (family F013) with progressive autosomal dominant hearing loss was mapped to a critical region spanning 28.54 Mb on chromosome 9q31.3-q34.3 by linkage analysis, which was a novel DFNA locus, assigned as DFNA56. In this interval, there were 398 annotated genes. Then, whole exome sequencing was applied in three patients and one normal individual from this family. Six single nucleotide variants and two indels were found co-segregated with the phenotypes. Then using mass spectrum (Sequenom, Inc.) to rank the eight sites, we found only the TNC gene be co-segregated with hearing loss in 53 subjects of F013. And this missense mutation (c.5317G>A, p.V1773M ) of TNC located exactly in the critical linked interval. Further screening to the coding region of this gene in 587 subjects with nonsyndromic hearing loss (NSHL) found a second missense mutation, c.5368A>T (p. T1796S), co-segregating with phenotype in the other family. These two mutations located in the conserved region of TNC and were absent in the 387 normal hearing individuals of matched geographical ancestry. Functional effects of the two mutations were predicted using SIFT and both mutations were deleterious. All these results supported that TNC may be the causal gene for the hearing loss inherited in these families. TNC encodes tenascin-C, a member of the extracellular matrix (ECM), is present in the basilar membrane (BM), and the osseous spiral lamina of the cochlea. It plays an important role in cochlear development. The up-regulated expression of TNC gene in tissue repair and neural regeneration was seen in human and zebrafish, and in sensory receptor recovery in the vestibular organ after ototoxic injury in birds. Then the absence of normal tenascin-C was supposed to cause irreversible injuries in cochlea and caused hearing loss.

MYO15A variants are responsible for human non-syndromic autosomal recessive deafness (DFNB3). The majority of MYO15A variants are associated with a congenital severe-to-profound hearing loss phenotype, except for MYO15A variants in exon 2, which cause a milder auditory phenotype, suggesting a genotype-phenotype correlation of MYO15A. However, MYO15A variants not in exon 2 related to a milder phenotype have also been reported, indicating that the genotype-phenotype correlation of MYO15A is complicated. This study aimed to provide more cases of MYO15A variation with diverse phenotypes to analyse this complex correlation. Fifteen Chinese autosomal recessive non-syndromic hearing loss (ARNSHL) individuals with MYO15A variants (8 males and 7 females) from 14 unrelated families, identified by targeted gene capture of 127 known candidate deafness genes, were recruited. Additionally, we conducted a review of the literature to further analyses all reported MYO15A genotype-phenotype relationships worldwide. We identified 16 novel variants and 12 reported pathogenic MYO15A variants in 15 patients, two of which presented with a milder phenotype. Interestingly, one of these cases carried two reported pathogenic variants in exon 2, while the other carried two novel variants not in exon 2. Based on our literature review, MYO15A genotype-phenotype correlation analysis showed that almost all domains were reported to be correlated with a milder phenotype. However, variants in the N-terminal domain were more likely to cause a milder phenotype. Using next-generation sequencing (NGS), we also found that the number of known MYO15A variants with milder phenotypes in Southeast Asia has increased in recent years. Our work extended the MYO15A variant spectrum, enriched our knowledge of auditory phenotypes, and tried to explore the genotype-phenotype correlation in different populations in order to investigate the cause of the complex MYO15A genotype-phenotype correlation.