The blockade of programmed cell death protein 1 (PD-1) as monotherapy has been widely used in melanoma, but to identify melanoma patients with survival benefit from anti-PD-1 monotherapy is still a big challenge. There is an urgent need for prognostic signatures improving the prediction of immunotherapy responses of these patients. We analyzed transcriptomic data of pre-treatment tumor biopsies and clinical profiles in advanced melanoma patients receiving only anti-PD-1 monotherapy (nivolumab or pembrolizumab) from the PRJNA356761 and PRJEB23709 data sets as the training and validation cohort, respectively. Weighted gene co-expression network analysis was used to identify the key module, then least absolute shrinkage and selection operator was conducted to determine prognostic-related long noncoding RNAs (lncRNAs). Subsequently, the differentially expressed genes between different clusters were identified, and their function and pathway annotation were performed. In this investigation, 92 melanoma patients with complete survival information (51 from training cohort and 41 from validation cohort) were included in our analyses. We initiallyidentified the key module (skyblue) by weighted gene co-expression network analysis, and then identified a 15 predictive lncRNAs (AC010904.2, LINC01126, AC012360.1, AC024933.1, AL442128.2, AC022211.4, AC022211.2, AC127496.5, NARF-AS1, AP000919.3, AP005329.2, AC023983.1, AC023983.2, AC139100.1, and AC012615.4) signature in melanoma patients treated with anti-PD-1 monotherapy by least absolute shrinkage and selection operator in the training cohort. These results were then validated in the validation cohort. Finally, enrichment analysis showed that the functions of differentially expressed genes between two consensus clusters were mainly related to the immune process and treatment. In summary, the 15 lncRNAs signature is a novel effective predictor for prognosis in advanced melanoma patients treated with anti-PD-1 monotherapy.

Background: The rapid integration of artificial intelligence (AI) and medical big data into health care is transforming diagnosis, treatment planning, and research. However, formal education in these areas remains limited in undergraduate medical curricula, particularly in China. Objective: This study aimed to investigate clinical medicine undergraduates’ familiarity with AI and medical big data, their perceived need for related courses, and their preferred curriculum design and assessment methods. Methods: A cross-sectional, web-based survey was conducted at Zunyi Medical University, Guizhou, China, from January 10 to 17, 2025. In the institutional context of this study, “clinical medicine” included related clinical-track specialties such as pediatrics and psychiatry. All eligible students (N=1094) were invited, and 871 (79.6%) were included in the final analysis. The self-administered questionnaire was developed based on a literature review and expert consultation, with content validity quantified using the content validity index. Descriptive statistics were used to summarize response distributions. For ordinal outcomes (items 1-14), adjusted ordinal logistic regression models were applied, with gender and grade as predictors and major as a covariate. Given the small number of third- and fourth-year students, grade was modeled as an ordered trend variable. For nominal outcomes (items 15-16), group differences were assessed using chi-square tests or Fisher exact tests, as appropriate. Results: A total of 871 students were analyzed, of whom 62.6% (n=545) were women. Overall familiarity with AI and medical big data was limited: 34.8% (303/871) agreed or strongly agreed that they were familiar with the topic, and only 33% (287/871) reported having at least some prior learning experience. In contrast, the perceived educational need was high: 94% (819/871) considered such a course at least somewhat necessary, 57% (497/871) reported that the course was needed or very needed, 75.5% (658/871) indicated that they would likely or definitely enroll, and 56.5% (492/871) reported that they would likely or definitely engage in self-directed learning. Personalized teaching based on textbooks (566/871, 65%) or open-book examinations (633/871, 72.7%) was the most preferred instructional and assessment format. Preferences for course materials and assessment methods differed by grade but not by gender. Conclusions: Early-stage clinical medicine undergraduates demonstrated limited familiarity with AI and medical big data but expressed a strong demand for related education. Students preferred structured yet flexible instructional formats and open-book assessments. Although the findings are based predominantly on first- and second-year students, they support the development of staged, practice-oriented AI and medical big data curricula tailored to the needs of early-stage clinical medicine undergraduates.