Proteomics in the age of artificial intelligence

Abstract

Proteomics is now generating unprecedented volumes of mass spectrometry data, driven by increasingly powerful instrumentation and large-scale studies. Yet our ability to translate these measurements into biological knowledge remains a major challenge. The limiting factor is no longer data generation, but our capacity to extract meaningful insights from increasingly complex molecular datasets. Artificial intelligence (AI) is beginning to reshape every stage of the proteomics workflow. Early applications focused on improving established computational approaches, where machine learning models enhanced conventional database search strategies by predicting peptide properties and improving peptide-spectrum match scoring. More recently, deep learning models are learning directly from tandem mass spectra for de novo peptide sequencing, enabling new discoveries. The next evolution will be AI-driven proteomics experiments, where models actively contribute to experimental decision-making. Rather than simply analyzing measurements after acquisition, AI could guide instruments toward collecting the most informative data, transforming proteomics from a strategy of measuring more toward one of measuring smarter. Ultimately, the greatest opportunity lies in moving beyond individual spectra and molecular identifications toward biological understanding. By integrating diverse omics measurements, AI may enable predictive representations of cellular systems and provide new ways to understand the complexity of life. I will explore the ongoing transformation of proteomics through AI: from models that enhance existing workflows, through systems that directly interpret molecular signals, toward intelligent technologies that could redefine how we perform experiments and uncover biological knowledge.