Integrating environmental factors and genetic variants in machine learning to assess occupational noise impact on health

Noise-induced hearing loss (NIHL) is a complex disorder arising from the interplay between noise, as well as contributions from other environmental and genetic risk factors. Although occupational noise exposure is a well-established risk factor, the extent to which other variables contribute remains poorly understood. This study aimed to investigate the nonlinear relationships among these variables using several machine learning algorithms, and to evaluate the relative contributions of environmental and genetic factors to the development of occupational NIHL. Data were collected from 2077 shipyard workers between 2012 and 2021. Noise exposure was quantified as cumulative noise exposure (CNE), estimated from workplace noise measurements and individual career duration. Genetic factors were identified through whole-exome sequencing-based association analyses and SNaPshot genotyping. The impact of co-variables, including sex, age, smoking status, and alcohol consumption were addressed. The classification model achieved 86 % accuracy (area under the curve [AUC]=0.80). Ranking analysis and logistic regression indicated that long-term equivalent (Leq) noise level had the strongest association with occupational NIHL (Leq noise level 80-85 dBA: odds ratio [OR]=7.04; 95 % confidence interval [CI], 2.37-20.89, p < 0.001; Leq noise level 90-95 dBA: odds ratio [OR]=10.87; 95 % confidence interval [CI], 1.42-82.97, p = 0.021), followed by age (OR=1.02; 95 % CI, 1.00-1.03, p = 0.028) and the CNPY2 rs10783780 single nucleotide polymorphism. The G allele of CNPY2, which regulates endoplasmic reticulum stress and cell survival, was associated with increased risk of NIHL (OR=1.39; 95 % CI, 1.16-1.67, p < 0.001). NIHL severity is significantly influenced not only by CNE and age, but also by mutations in the CNPY2 gene, which regulates endoplasmic reticulum stress and cell survival. These findings suggest that effective prevention of occupational NIHL should encompass not only noise control measures to reduce CNE, but also consideration of individual factors such as age and genetic susceptibility, including CNPY2 variants.