Machine learning-driven discovery of quinoline derivatives against Ichthyophthirius multifiliis

Ichthyophthirius multifiliis , a ciliate pathogen causing White Spot Disease in freshwater aquaculture, is responsible for substantial economic losses globally. This study establishes a machine learning (ML)-driven workflow for the efficient discovery of novel antiparasitic agents. The antiparasitic efficacy of 24 quinoline derivatives was evaluated against the theront and tomont life stages. Halogenated 8-hydroxyquinolines were identified as the most potent class, with 5,7-dichloro-2-methyl-8-quinolinol (Compound 13) emerging as the lead compound, demonstrating EC₅₀ values of 0.271 mg/L against theronts and 0.184 mg/L against tomonts. A predictive Quantitative Structure-Activity Relationship (QSAR) framework was established, with the Random Forest (RF) model showing superior predictive performance (test set R 2  = 0.761 for theronts and 0.703 for tomonts). Subsequently, the validated model was applied to screen a virtual compound library, identifying four novel quinoline candidates. Experimental validation of these ML-prioritized compounds confirmed their high potency, demonstrating the model's robust capability to identify new active scaffolds beyond the training set chemical space. Interpretation of molecular descriptors revealed that halogen-mediated electronic effects, molecular topological symmetry, and balanced hydrophobicity are critical factors governing antiparasitic activity. This work not only identifies highly promising candidates for the control of White Spot Disease but also provides a scalable, ML-accelerated paradigm for antiparasitic drug discovery in aquaculture.