Harnessing machine learning and multi-omics to explore tumor evolutionary characteristics and the role of AMOTL1 in prostate cancer.

Although recent advancements have shed light on the crucial role of coordinated evolution among cell subpopulations in influencing disease progression, the full potential of these insights has not yet been fully harnessed in the clinical application of personalized precision medicine for prostate cancer (PCa). In this study, we utilized single-cell sequencing to identify the evolutionary characteristics of tumoral cell states and employed comprehensive bulk RNA sequencing to evaluate their potential as prognostic indicators and therapeutic targets. Leveraging advancements in artificial intelligence, we integrated machine learning with multi-omics to develop and validate the tumor evolutionary characteristic predictive indicator (TECPI). TECPI not only demonstrated superior prognostic performance compared to traditional clinical predictors and 81 previously published models but also improved patient outcomes by accurately identifying individuals who would benefit from immunotherapy and targeted therapies. Furthermore, we experimentally validated the critical role of AMOTL1 in PCa pharmacodynamics through its interaction with AR, pivotal for modulating the sensitivity to AR antagonist. Additionally, we demonstrated the generalizability and applicability of TECPI across pan-cancers. In summary, this study emphasizes the importance of understanding cellular diversity and dynamics within the tumor microenvironment to predict PCa progression and to guide targeted therapy effectively.