The convergence of mass spectrometry (MS) and modern artificial intelligence (AI) offers transformative advantages, and a paradigm shift in the way we extract and interpret complex information from MS datasets. This presentation will demonstrate how AI approaches, particularly computer vision, generative AI and AI agents, can democratize access to sophisticated MS-based omics analyses while simultaneously enhancing accuracy, breadth of coverage, and holistic processing and interpretation across multiple omics technologies.
Covering various proteomics, metabolomics, and lipidomics datasets from biological, environmental and synthetic biology studies, I will showcase three innovative tools that we have developed to address current limitations in MS data processing and interpretation. The first two are omics-agnostic tools. PeakQC provides user-friendly, accessible and comprehensive quality control for raw MS data from various instrumentations. PeakDecoder enables sophisticated molecular annotation and quantitation of multidimensional datasets from analyses with LC-ion mobility-MS and data-independent acquisition. Our third tool, PTMdiscoverer, represents an alternative approach to find post-translational modifications (PTMs) by identifying delta masses in peptide fragmentation patterns from LC-MS proteomics data, and utilizing a large language model for their annotation and interpretation across various phenotypes or experimental conditions.
Finally, I will outline my vision for integrating our current tools and future developments into scalable agentic AI systems capable of orchestrating automated next-generation multiomics analysis. Our approaches strive to accelerate bio-discoveries by enabling researchers, regardless of their level of computational expertise, to extract meaningful biological insights from complex MS datasets and, ultimately, advance our understanding of biological processes across scientific disciplines.