Cancer Cell：邵志敏/江一舟团队等利用AI辅助分型，提高乳腺癌治疗效果

Core Viewpoint - The study highlights the potential of AI-assisted molecular subtyping to enhance treatment efficacy in HR+/HER2- breast cancer, particularly for the newly identified subtypes SNF2 and SNF4 [4][22]. Group 1: Research Background - HR+/HER2- breast cancer accounts for approximately two-thirds of all breast cancer cases, with limited treatment options available after resistance to CDK4/6 inhibitors [3][4]. - The study published in Cancer Cell on December 4, 2025, by researchers from Fudan University, introduces a new approach to classify HR+/HER2- breast cancer into four subtypes using AI [3][4]. Group 2: AI Subtyping Methodology - The research team utilized multi-omics data and a similarity network fusion (SNF) algorithm to classify HR+/HER2- breast cancer into four new subtypes: SNF1, SNF2, SNF3, and SNF4 [8][11]. - An AI model was developed to predict these subtypes accurately from routine H&E stained pathology slides, making clinical application feasible without expensive multi-omics testing [11][21]. Group 3: LINUX Trial Design - The LINUX trial is a multi-center, randomized controlled Phase II study involving 105 HR+/HER2- advanced breast cancer patients who are resistant to CDK4/6 inhibitors [13][14]. - Patients were categorized based on their SNF subtype and randomly assigned to either a precision treatment group or a standard treatment group [13][14]. Group 4: Treatment Protocols - Treatment regimens were tailored to each subtype: - SNF1: Everolimus + endocrine therapy - SNF2: Camrelizumab + Famitinib + chemotherapy - SNF3: Olaparib + chemotherapy - SNF4: Apatinib + chemotherapy [14]. Group 5: Trial Results - The trial demonstrated significant efficacy, with objective response rates (ORR) for the precision treatment group compared to the control group as follows: - SNF1: 10% vs 0% - SNF2: 65% vs 30% - SNF3: 40% vs 30% - SNF4: 70% vs 20% [17]. - Notably, the probability of effective treatment for SNF2 and SNF4 was 86.7% and 97.6%, respectively, indicating successful patient selection for specific therapies [19]. Group 6: Safety and Clinical Implications - The safety profile of the precision treatment group was comparable to the standard treatment group, with a 37% incidence of grade 3-4 treatment-related adverse events [19]. - The LINUX trial results provide strong evidence for new treatment strategies in HR+/HER2- advanced breast cancer, particularly identifying SNF2 as an "immune hot tumor" type [22][24]. Group 7: Future Outlook - The success of the LINUX trial validates the feasibility of using low-cost pathology and AI for clinical precision subtyping and treatment guidance [21]. - This research paves the way for personalized treatment approaches in breast cancer, aiming for a "one patient, one plan" strategy in precision medicine [24].