Background: The improvement of global health literacy depends on health professionals having effective public science communication capabilities. However, worldwide, medical education systems still largely lack training in science communication for medical students. How to construct a systematic and scalable model for cultivating scientific communication capabilities is a key issue currently faced by the reform of medical education. Objective: Our study aims to evaluate the effect of the “5E+AI” teaching model on the cultivation of medical students' science communication capabilities, and to assess the impact of this model on students' learning participation, acquisition of medical knowledge, and creation of popular science works. Methods: A historical controlled quasi-experimental design was employed. Twenty-nine medical students enrolled in the “Fundamentals and Practice of Medical Science Popularization” course at Sun Yat-sen University during the 2025-1 semester served as the historical control group (receiving lecture-based, instructor-centered teaching with PowerPoint presentations and group discussions, but without AI-assisted tools), while 28 students enrolled in the same course during the 2025-2 semester served as the experimental group (receiving the “5E+AI” teaching model). The two groups are the same in terms of teaching content, total class hours, teachers, teaching materials, assessment methods, etc. The evaluation indicators include participation score, medical knowledge test score, final score of the science communication project and student self-report questionnaire. Intergroup comparisons were performed using the Mann–Whitney U -test. Results: The experimental group scored significantly higher than the control group on all three primary outcome measures: participation scores (median 85.00 vs. 80.00, U = 220.00, P = 0.002), medical knowledge test scores (median 100.00 vs. 80.00, U = 210.00, P = 0.001), and science communication project scores (median 85.00 vs. 85.00, U = 213.00, P = 0.001). In addition, the experimental group reported significantly higher scores than the control group across all three dimensions of the self-report questionnaire ( P < 0.05). Conclusions: The “5E+AI” teaching model helps enhance medical students' classroom participation, ability to acquire medical knowledge and science communication skills. In this teaching model, artificial intelligence can help students address the issues of knowledge deficiency and resource constraints. Meanwhile, the 5E teaching framework provides students with structured teaching scenarios. Together, they enhance teaching effectiveness and offer replicable and adaptable local solutions for the reform of medical communication education in the context of global health literacy promotion.

Background: Oncology education often focuses on disease burden and mortality, which may reduce students’ learning interest and reinforce negative perceptions of cancer. Although positive psychology has shown educational benefits in promoting motivation, resilience, and professional development, its integration into oncology education remains insufficiently explored. Objective: To evaluate the effectiveness of a positive psychology–integrated oncology teaching model (4C-HOPE) in undergraduate medical education. Methods: Sixty-eight undergraduate medical students at Sun Yat-sen University participated in the 4C-HOPE teaching intervention during the hepatocellular carcinoma (HCC) chapter of their oncology course. Knowledge tests and questionnaires were administered before and after the intervention to assess educational outcomes. Student satisfaction and subsequent clinical clerkship performance were also evaluated. Results: Knowledge test scores improved significantly, with the median score increasing from 60 to 100 ( P < 0.001). The proportion of students reporting strong interest in learning about HCC increased from 50.0% to 85.0% ( P < 0.001). Agreement that physicians should provide holistic care increased from 89.7% to 95.0% ( P = 0.480), while willingness to apply positive psychology-based communication skills increased from 60.3% to 77.5% ( P = 0.089). Overall, 87.5% of students expressed high satisfaction with the teaching model. During subsequent clerkships, students exposed to the 4C-HOPE model achieved higher examination scores than controls (81.49 ± 7.87 vs. 77.45 ± 8.58, P = 0.014). Conclusions: The 4C-HOPE teaching model significantly improved knowledge acquisition and learning interest among undergraduate medical students, with favorable trends in holistic care awareness and communication willingness. The sustained benefits observed in clerkship performance further support its pedagogical value. This model offers a feasible framework for fostering both academic learning and professional development in oncology education.

Background: Hepatocellular carcinoma (HCC) remains a major cause of cancer-related mortality, and responses to immune checkpoint blockade are often limited. Casein kinase 2 beta (CK2beta), encoded by CSNK2B, has established tumor-intrinsic functions, but its relationship with immune states in HCC remains unclear. Methods: We integrated TCGA-LIHC, six GEO cohort-platform combinations, GSE149614 single-cell RNA sequencing data, CIBERSORTx deconvolution, broad-compartment CellChat inference, and public proteomic resources. We evaluated CSNK2B expression, overall survival, a prespecified GDF15-associated Treg/checkpoint transcriptional module, tumor microenvironment features, compartment-level localization, and protein-level support. Results: CSNK2B was upregulated in TCGA-LIHC and all six external cohort-platform comparisons but was not a robust prognostic marker. The strongest immune association was with the GDF15-associated Treg/checkpoint module (Spearman rho = 0.371, q = 2.9 × 10⁻¹²); the direction was positive in all six GEO analyses and four passed BH-FDR. Deconvolution results were small and heterogeneous. CSNK2B and GDF15 were most prominent in hepatocyte-like compartments, while the candidate hepatocyte-to-T/NK GDF15–TGFBR2 pair was not reproduced in pooled data or in any of 10 patient-stratified CellChat runs. Public proteomic datasets supported tumor-side elevation of both proteins. Conclusions: CSNK2B expression aligns with a reproducible GDF15-associated transcriptional state in HCC, accompanied by selective stromal, endothelial, and immune-context changes. The evidence is associative and provides hypotheses for perturbation, spatial, and co-culture validation rather than a causal or treatment-prediction claim.