Abstract
Early detection and intervention of precancerous lesions are crucial in reducing cancer morbidity and mortality. Comprehensive analysis of genomic, transcriptomic, proteomic and epigenomic alterations can provide insights into the early stages of carcinogenesis. However, the lacke of an integrated, well-curated data resource of molecular signatures limits our understanding of precancerous processes. Here, we introduce a comprehensive PreCancerous Molecular Resource (PCMR), which compiles 25,828 molecular profiles of precancerous samples paired with normal or malignant counterparts. These profiles cover precancerous lesions of 35 cancer types across 20 organs and tissues, derived from tissue samples, liquid biopsies, cell lines and organoids, with data from transcriptomics, proteomics and epigenomics. PCMR includes 62,566 precancer-gene associations derived from differential analysis and text-mining using the ChatGPT large language model. We examined PCMR dataset reliability and significance by the authoritative precancerous molecular signature, along with its biological and clinical relevance. Overall, PCMR will serve as a valuable resource for advancing precancer research and ultimately improving patient outcomes.