Multivariate analysis of a large-scale auditory screen reveals new candidate hearing loss genes

Early-onset hearing loss (HL) occurs in ~ 1 in 500 babies, with genetic factors thought to account for ~ 80% of prelingual HL cases. Hereditary HL is highly heterogenous, with > 100 genes identified as critical for hearing. However, most patients remain undiagnosed after genetic testing, suggesting that many more hearing essential genes are yet to be identified. The International Mouse Phenotyping Consortium (IMPC) provides an unparalleled resource for ‘hearing’ gene discovery, with > 6,700 mouse knockouts assessed by auditory brainstem response (ABR) testing. However, the current frequency-by-frequency univariate analyses applied to ABR measurements limit the sensitivity and specificity of current IMPC auditory phenotyping. Here, integrating multivariate, Bayesian, and clustering approaches to interrogate the IMPC datasets, we identify 133 genes with evidence for HL, of which 59 are novel associations. Importantly, 16 of the 59 genes are not identified by the statistical pipeline used by the IMPC. We also apply unsupervised clustering to derive data-driven audiometric phenotypes, revealing recurrent patterns suggestive of shared underlying mechanisms. Moreover, we validate two candidate genes, Tmem51 and Nedd4l . Our analyses demonstrate the IMPC remains a valuable resource for the identification of novel deafness genes and provide a methodological framework that could be applied across other large-scale phenotyping pipelines.