Non-lethal molecular biomonitoring of pyrethroid exposure in freshwater fish using FTIR spectroscopy and machine learning.

Pyrethroid pesticides such as lambda-cyhalothrin (LC) continue to pose substantial risks to freshwater ecosystems, where sublethal exposures may impair physiological function in non-target organisms. However, conventional monitoring techniques remain limited in their ability to detect early biochemical alterations, particularly in native fish species. Here, we demonstrate that Fourier Transform Infrared (FTIR) spectroscopy, combined with multivariate analysis and supervised machine learning, provides a rapid and non-destructive approach to detect LC exposure in Astyanax lacustris fin tissue. FTIR revealed distinct molecular signatures associated with lipid peroxidation and protein structural modifications, accompanied by increased intra-group biochemical heterogeneity following exposure. Principal component analysis (PCA) discriminated exposed and control specimens, and a support vector machine (SVM) classifier trained on principal component scores achieved 91% accuracy in external validation. These results show that fin tissue, which is accessible through minimally invasive collection, can serve as an effective sentinel substrate for molecular biomonitoring. The combined FTIR-machine learning framework represents a sensitive and operationally simple strategy for detecting pesticide impacts in freshwater systems. This approach has potential to support future regulatory and ecological monitoring once validated under environmentally realistic conditions.