Liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS) is one of the most effective tools for identifying changes in plant metabolic profiles after stress. Studying plant metabolic response can, in fact, be exploited for developing strategies for early detection of pathogen infections, thus reducing the risk of contamination and safeguarding the safety of the food chain. In this work we explored the application of Machine Learning (ML) approaches for the discrimination of maize kernels infected by Fusarium verticillioides from safe (i.e. non-infected) kernels. In particular Machine Learning (ML) algorithms were applied to data (peak lists) obtained from LC-HRMS analysis. We therefore proceeded with the identification of inoculated (infected) and control samples from the peaks list comparing three well known machine learning models: XGBoost, Random Forest and a Feed Forward neural network based architecture. Preliminary experiments showed promising performances, reaching over 80% of accuracy in the detection of infected samples. This approach will be further improved to build models for the early detection of Fusarium infection.