Evaluating microplastics and antibiotics induced genotoxicity in marine mussels through deep learning-based processing images of comet assay

The increasing diversity of emerging pollutants poses significant risks to aquatic ecosystems, particularly in terms of genotoxicity. Monitoring genetic damage in model organisms can offer early warnings about environmental risks linked to emerging pollutants. The common use of the comet assay in genotoxicity necessitates the creation of high-throughput methods for broader application. The aim of this study was to develop a deep learning (DL)-based high-throughput comet assay analysis method for automated, accurate detection and evaluation of DNA damage. This research introduces a technique utilizing DL algorithms for automated analysis of image batches, from raw inputs to results. The DL method employs the You Only Look Once version 8 model for comet detection on a dataset with 1453 training images and 364 validation images, with 6749 manually annotated comet subjectives. The DL method achieves 90.6 % accuracy for comets with tails. Comparative analysis with traditional methods was conducted, showing the DL method's effectiveness in genotoxicity evaluation. Application of the DL method and Comet Assay Software Project (CASP) for assessing genotoxicity induced by combined stressors (polyethylene microplastics and sulfamethoxazole) in mussel gills revealed that while individual exposure to either stressor significantly increased DNA damage, their co-exposure exhibited a mitigating effect on overall genotoxicity. The DL method correlated strongly with the CASP method (R= 1.09, P < 0.01). This integrated approach combines classical toxicology with DL to provide an efficient and accurate way to assess DNA damage from emerging pollutants.