Early detection and intervention of cancer or precancerous lesions hold great promise to improve patient survival. However, the processes of cancer initiation and the normal–precancer–cancer progression within a non-cancerous tissue context remain poorly understood. This is, in part, due to the scarcity of early-stage clinical samples or suitable models to study early cancer. In this Review, we introduce clinical samples and model systems, such as autochthonous mice and organoid-derived or stem cell-derived models that allow longitudinal analysis of early cancer development. We also present the emerging techniques and computational tools that enhance our understanding of cancer initiation and early progression, including direct imaging, lineage tracing, single-cell and spatial multi-omics, and artificial intelligence models. Together, these models and techniques facilitate a more comprehensive understanding of the poorly characterized early malignant transformation cascade, holding great potential to unveil key drivers and early biomarkers for cancer development. Finally, we discuss how these new insights can potentially be translated into mechanism-based strategies for early cancer detection and prevention.

癌症或癌前病变的早期检测和干预对提高患者生存率具有很大的希望。然而，在非癌组织背景下，癌症发生的过程和正常的 - 癌前 - 癌症进展仍然知之甚少。这在一定程度上是由于缺乏早期临床样本或研究早期癌症的合适模型。在这篇综述中，我们介绍了临床样本和模型系统，例如本土小鼠和类器官衍生或干细胞衍生的模型，这些模型允许对早期癌症发展进行纵向分析。我们还介绍了增强我们对癌症发生和早期进展的理解的新兴技术和计算工具，包括直接成像、谱系追踪、单细胞和空间多组学以及人工智能模型。这些模型和技术共同促进了对特征不佳的早期恶性转化级联反应的更全面理解，具有揭示癌症发展的关键驱动因素和早期生物标志物的巨大潜力。最后，我们讨论了如何将这些新见解潜在地转化为基于机制的早期癌症检测和预防策略。