Premalignant lesions serve as critical transitional states in the development of many cancers. They also provide a scientific basis for the traditional Chinese medicine (TCM) concept of “treating diseases before they arise.” However, the progression of premalignant lesions is highly heterogeneous—characterized by long timelines, low transformation rates, and unclear molecular triggers. Consequently, predicting which individuals will develop cancer and identifying the earliest actionable window for intervention remain major challenges.
The team led by Prof. Shao Li at Tsinghua University has long focused on the mechanisms and early prevention of inflammation–cancer transformation. Their earlier work defined a newly recognized stage—the “exceedingly-early” stage of gastric cancer—and identified its cellular features, biomarkers, and therapeutic targets, forming an intelligent and precise early-prevention framework. These studies have appeared in Cancer Discovery, Cancer Research, Nature Communications, and Cell Reports.

Research Progress
Building upon this foundation, the research team integrated multimodal clinical information, histopathology, and multi-omics data with artificial intelligence to systematically analyze 15 types of premalignant lesions across multiple organs. Using biological network–based modeling, the researchers characterized pathological changes, progression risks, and dynamic molecular features associated with inflammation–cancer transitions.

This unified framework enables the reconstruction of disease evolution trajectories from premalignancy to the exceedingly-early stage, offering a cross-organ, cross-dataset perspective on early malignant transformation. The work establishes the concept of a ‘pan-cancer exceedingly-early’ window and provides computational principles for risk prediction, biomarker identification, and early intervention. The study, titled “Multi-Omics Meets Premalignancy: Pave the Way for Cancer Early Prevention,” has been published in Research (IF 10.7).

Future Prospects
The ‘pan-cancer exceedingly-early’ strategy provides a promising avenue for intelligent early cancer prevention. By combining multimodal data integration, biological networks, AI modeling, and insights from Chinese and Western medicine, this framework supports systematic early warning and personalized early intervention before malignant transformation occurs.
As computational and clinical technologies continue to advance, this approach may contribute to establishing scalable early-prevention systems across diverse cancer types and promote the development of Chinese-original early cancer prevention pathways.

The complete study is accessible via DOI: 10.34133/research.0961