ALKBH5 orchestrates ferroptosis-driven tumor suppression: An LLM-powered discovery in prostate cancer.

The treatment of prostate cancer (PCa) continues to pose substantial clinical challenges. The use of large language models (LLMs) to identify the key molecular determinants of PCa progression, followed by experimental biological validation, helps uncover novel therapeutic targets. We developed hierarchical knowledge-guided LLM for risk gene identification (HKLLM-RG), a PCa risk gene identification method. Among the candidate genes identified, ALKBH5 emerged as particularly noteworthy in the analysis. Reduced expression of ALKBH5 correlated with aggressive clinical features and significantly reduced survival in PCa. ALKBH5 inhibits PCa progression and promotes ferroptosis. CHRM3, which is the downstream molecule of ALKBH5, could promote PCa cell proliferation and migration. ALKBH5 regulates CHRM3 in an mA-dependent manner. Mechanistically, the ALKBH5/CHRM3 axis suppresses AKT signaling, thereby inducing the upregulation of the transcriptional repressor ZNF281. This regulatory cascade subsequently downregulates the expression of SLC3A2 and GPX4, ultimately sensitizing cells to ferroptosis. Thus, AZD5363 and RSL3 targeting the ALKBH5/CHRM3/ZNF281 axis can effectively synergize to treat PCa by promoting ferroptosis. Taken together, this study leverages LLM-guided discovery to delineate a novel ALKBH5/CHRM3/ZNF281 regulatory axis controlling ferroptotic susceptibility in PCa. Importantly, a synergistic therapeutic strategy was identified by combining RSL3 with AZD5363, providing novel therapeutic targets and directions for PCa treatment.