WormSpot: a machine learning-powered viability scoring platform in C. elegans for Candida pathogenicity studies

Abstract

Invasive Candida infections pose a critical health challenge, exacerbated by emerging antifungal resistance. Caenorhabditis elegans (C. elegans) offers a genetically tractable and scalable model for studying Candida pathogenicity, yet conventional viability assays remain labor-intensive, limiting high-throughput applications. In this study, we developed a machine learning-driven worm viability scoring platform, WormSpot, using representative images of Candida-infected worm and the robust YOLOv8-based framework. By analyzing static morphological features, our model accurately classifies worm viability post-infection, achieving strong concordance with manual scoring while requiring minimal image input. Validation with well-characterized Candida albicans mutants and antifungal agents confirms the platforms robustness in predicting worm survival trends. Notably, WormSpot performs efficiently with diverse image formats and is compatible with standard multi-well microscopy workflows, enabling automated data analysis and scalable application in various experimental setups. In conclusion, WormSpot provides a data-efficient, reproducible tool for assessing Candida virulence using C. elegans, supporting both basic pathogenicity study and antifungal discovery in high-throughput settings.