Fungal infection caused by invasive Aspergillus is a life-threatening complication in immunocompromised pediatric cancer patients. However, the early diagnosis of invasive infection remains a clinical challenge due to the lack of specific, non-invasive biomarkers. The current study investigates plasma metabolomic profiling integrated with an AI-derived fungal secondary metabolite database to identify potential biomarkers for rapid, non-invasive detection of Aspergillus infection. Plasma samples from thirteen pediatric oncology patients were analyzed using untargeted metabolomics based on UHPLC-MS/MS. Based on galactomannan assay results, three patients were classified as Aspergillus-Infected (AIC) and ten as non-infected controls (NPCs). An in-house custom database for secondary metabolites of fungi was incorporated to enhance metabolite annotation. Eight metabolites were found to be candidate biomarkers based on statistical significance, fold change, and biological relevance. In the AIC cohort, aflatoxin B1, aspergillimide, fumifungin, and uridine were found to be significantly elevated while citric acid presented a decrease. Multivariate analysis utilizing PCA and PLSDA showed distinct group separation. Moreover, sample size estimation indicates that a minimum of 25 participants would be needed in future studies for appropriate statistical power.