Abstract

Entomopathogenic nematodes (EPNs) are widely used in biological pest control, yet their efficacy in field applications is highly dependent on physiological quality during storage; an area lacking direct, predictive quality control tools. Current assessments rely solely on labor-intensive infectivity tests using insect hosts like Galleria mellonella. This study breaks new ground by combining Fourier Transform Infrared Spectroscopy (FTIR-ATR), two-dimensional correlation spectroscopy (2D-COS), and machine learning to deliver a rapid, non-destructive method for evaluating EPNs viability under thermal stress. This study examines the biochemical responses of Heterorhabditis indica and Steinernema riobrave subjected to thermal stress at 10 degrees C, 20 degrees C, and 30 degrees C over an 8 week. Infectivity responses varied significantly between species under thermal stress. S. riobrave declined substantially after six weeks at 30 degrees C (-55 %), whereas H. indica retained considerably higher infectivity at the same temperature, demonstrating superior thermal tolerance. FTIR spectra revealed temperature- and speciesdependent changes across lipid, carbohydrate, and protein spectral regions, indicating distinct biochemical adaptation strategies. Glycogen and triglyceride levels dropped sharply in S. riobrave at 30 degrees C (up to -65 %), while H. indica exhibited elevated trehalose accumulation, particularly at 20 degrees C, suggesting enhanced protective mechanisms. Machine learning models showed strong predictive performance., with Support Vector Machine algorithms trained on spectral data accurately predicting infectivity (R2 = 0.93, RMSE = 0.04). Furthermore, 2DCOS analysis highlighted trehalose and protein amide I spectral shifts as early diagnostic markers of thermal stress, preceding observable infectivity decline. Together, these findings introduce a novel, integrative platform for EPNs quality control that is rapid, sensitive, and predictive. This approach offers a transformative advance for commercial biocontrol production, allowing early detection of stress-induced viability loss and optimization of storage protocols to ensure field performance.