Untargeted metabolomics is a powerful tool for investigating chemistry of complex biological systems, but its utility is compromised by the presence of uninformative features and the limited efficiency of currently available prioritization tools. More effective filtering and prioritization tools are required to address the challenges of large untargeted metabolomics datasets. Here, we introduce Metabolomics Peak Analysis Computational Tool (MPACT), a new mass spectrometry data analysis platform employing filtering based on multiple modalities, statistical techniques incorporating multilevel replication, and interactive data visualization. We demonstrate application of MPACT to uncover hidden effects of the rare earth element cerium on tunicate-associated bacterium Streptomyces sp. PTY087I2, culminating in characterization of two thiolated compounds including a new cysteine derivative, granaticin C, and granaticin D, recently described as mycothiogranaticin A. While we demonstrate application of MPACT to microbial natural products discovery using an elicitation approach, the platform should be readily adaptable to investigation of multipartite interactions, biomarker detection, small molecules in the environment, and a wide range of other complex sample types.