Queen Mary Hospital

Oral cancer often develops from oral potentially malignant disorders. Oral leukoplakia (OL) is the most common oral potentially malignant disorder. However, not all patients with OL develop oral cancer in their lifetimes, and cancer risk assessment is challenging. This study developed a time-to-event artificial intelligence-based model (termed OralCancerPredict) that integrated patient characteristics, histologic features, and immunohistochemical indicators for precise prediction of cancer progression in OL. We performed a retrospective analysis of patient data (n = 456) and tissue samples (n = 1,312) obtained before cancer progression or at last follow-up for patients with OL. KRT13 and p53 immunohistochemistry and quantitative analyses were performed for tissue samples, while independent prognostic features, including demographic characteristics, clinical information, and histopathologic findings, were obtained for patients in the cohort. KRT13 and p53 expression was combined with histologic and patient data for model training, testing, and external validation. Performance evaluation for OralCancerPredict included model discriminative ability, calibration, explainability, and net benefit analysis via decision curves. External validation was also performed to ensure model generalizability based on patient data (n = 119) and tissue samples (n = 322) unused for training and testing. Our findings showed that OralCancerPredict trained on multidimensional data had good concordance indices (0.855 to 0.867), areas under the curve (0.877 to 0.882), and integrated Brier scores (0.046 to 0.069) at testing and external validation. Explainability analysis confirmed the importance of KRT13 and p53 deregulated expression and World Health Organization dysplasia grading to OralCancerPredict in predicting the risk of cancerization and cancer progression-free survival. Moreover, the model had a better net benefit than the World Health Organization and binary dysplasia grading systems alone, which represent the current risk stratification methods in OL management. Overall, OralCancerPredict can predict the risk of cancer development and cancer progression-free survival for patients with OL, with good discrimination, calibration, and net benefit. The explainable artificial intelligence model has the potential to streamline intervention and close monitoring in the clinical management of OL.

Abstract Background Echocardiography (echo) notes contain valuable prognostic information for patients in the intensive care unit (ICU). However, their unstructured format and the presence of sensitive patient information present challenges for large-scale, automated analysis. There is a need for secure and efficient methods to extract and utilize echo data to enhance ICU outcome prediction. Methods We developed an AI-powered, privacy-preserving pipeline that leverages advanced natural language processing and pattern matching to annotate echo notes locally, ensuring comprehensive masking of personally identifiable information. This pipeline was applied to patient data from a mixed medical surgical ICU in a tertiary referral hospital. Key variables were extracted from echo notes and integrated with clinical and laboratory data to predict ICU mortality. A LightGBM machine learning model—robust to missing values—was trained using both routine and echo-derived structured clinical features. Its predictive performance was compared to that of the APACHE IV score. Results Compared with the reference standard derived from manual annotation by echocardiography specialist, automated annotation of echo notes achieved 98.85% data accuracy with a false positive rate of 0.31%. Several echo-derived variables, including left ventricular ejection fraction (LVEF), left ventricular outflow tract velocity time integral (LVOT VTI), tricuspid annular plane systolic excursion (TAPSE), mitral regurgitation (MR), and aortic regurgitation (AR), were strongly associated with ICU mortality. Incorporating echo-derived variables improved the accuracy in prediction of ICU mortality, with the LightGBM model achieving an AUC of 0.902 compared to 0.861 for APACHE IV score. Conclusion Our locally deployable AI pipeline enables secure and automated extraction of prognostic information from echo notes, substantially enhancing ICU mortality prediction. The inclusion of echo-derived variables significantly improved predictive accuracy, underscoring the potential but currently underutilized value of unstructured notes. This approach paves the way for scalable, privacy-preserving decision support tools in critical care.

The orally available toll-like receptor 7 (TLR7) agonist prodrug RO7119929 is converted to active drug predominantly in the liver, where it is hypothesized to reprogram the local immune microenvironment. We aimed to explore the safety, pharmacokinetics (PK), and preliminary antitumor activity and obtain the proof-of-mechanism for RO7119929 in patients with liver cancer. RO7119929 was investigated in mouse tumors and liver tissue from cynomolgus monkeys. In the first-in-human, open-label, dose-escalation and expansion study, patients with histologically confirmed advanced or metastatic primary liver cancers or solid tumors with predominant liver involvement received RO7119929 weekly in 3-week cycles either on a flat-dose (FD) or step-up dose (SUD) schedule. Preclinical results demonstrated antitumor activity and the proinflammatory proof-of-mechanism in mouse liver tumors and cynomolgus monkey liver tissue. The subsequent first-in-human study enrolled 27 patients in five FD dose-escalation cohorts, 18 patients in a FD dose-expansion cohort, and 9 patients in two SUD dose-escalation cohorts. The most common primary tumor type at study entry was hepatocellular carcinoma (HCC) (31%). PK data showed fast conversion of RO7119929 to the active TLR7 agonist. Treatment-related adverse events (TRAEs) were observed in 50 (91%) patients across all treated patients; the most commonly reported TRAE was cytokine release syndrome (CRS) in 48 (81%) patients. CRS was also identified as a dose-limiting safety risk and had a dose-dependent incidence and severity. Grade 3 CRS occurred in six (13%) patients receiving FD RO7119929, while no grade 3 CRS was reported in SUD patients at comparable target dose levels. TLR7-related transient, dose-dependent gene expression and cytokine induction was observed both with FD and SUD treatment and corresponded with treatment-induced local inflammation of the tumor microenvironment induced by reprogramming of the myeloid cells. Among 17 patients with HCC, one durable complete response was observed, and 10 (59%) patients had stable disease. SUD appears to reduce the risk of CRS while maintaining mode-of-action-relevant pharmacodynamic effects. The clinical activity of RO7119929 as a single agent was limited, suggesting that combination therapy with a checkpoint inhibitor may be needed to leverage the proinflammatory potential of RO7119929 and further increase antitumor activity. NCT04338685.

To determine the future priorities for ectopic pregnancy research. Potential research questions were collated from an initial international survey, a systematic review of clinical practice guidelines, and Cochrane systematic reviews. A rationalized list of confirmed research uncertainties was prioritized in an interim international survey. Prioritized research uncertainties were discussed during a consensus development meeting. Using a formal consensus development method, the modified nominal group technique, diverse stakeholders identified the top 10 research priorities for future ectopic pregnancy research. Healthcare professionals, people with lived experience of ectopic pregnancy, and others were brought together in an open and transparent process using formal consensus methods advocated by the James Lind Alliance. Not applicable. Top 10 research priorities for ectopic pregnancy. The initial survey was completed by 855 participants from 35 countries, and 1,220 potential research questions were submitted. Three clinical practice guidelines and 43 Cochrane systematic reviews identified a further 24 potential research questions. A rationalized list of 49 confirmed research uncertainties was entered into an interim prioritization survey completed by 413 respondents from 20 countries. The top 10 research priorities were identified during a consensus development meeting involving 37 participants from 10 countries. These research priorities are diverse and seek answers to questions regarding prevention, treatment, and the longer-term impact of ectopic pregnancy. We anticipate that the identified research priorities, developed to specifically highlight the most pressing clinical needs as perceived by healthcare professionals, people with lived experience of ectopic pregnancy, and others, will help research funding organizations and researchers to develop their future research agenda.

Abstract NRAS mutations drive 20–30% of melanomas and are associated with poor response to immunotherapies. RAS(ON) inhibitors such as daraxonrasib (RMC-6236) have shown promising activity in NRAS-mutant melanoma, yet mutation-specific sensitivity, resistance mechanisms, and rational combination strategies remain largely unresolved. Here, we performed saturation mutagenesis across NRAS mutation hotspots to systematically evaluate mutation-resolved oncogenic fitness and drug responsiveness to six direct RAS inhibitors (sotorasib, adagrasib, ADT-007, BI-2865, RMC-6236, and RMC-7977) using isogenic melanoma spheroids, xenografts, and pooled drug-sensitivity profiling. RMC-6236 and RMC-7977 displayed the broadest activity, sensitizing ∼80% of recurrent NRAS mutations across the P-loop and Switch II regions. Several clinically observed mutations, including Q61P, G13D/V/R, and G60E, showed reduced susceptibility to RAS(ON) inhibition. Structural modelling and thermostability analysis indicated impaired inhibitor engagement or reduced tricomplex stability as potential mechanisms of resistance. RAS(ON) blockade also induced STAT3 activation, particularly in oncogenic Q61 mutants. Co-targeting STAT3 with napabucasin or siRNA enhanced apoptosis, suppressed MYC, and improved tumour control in vitro and in vivo. Together, this study provides the first comprehensive mutation-resolved sensitivity atlas for NRAS-mutant melanoma, highlights structural mechanisms underlying resistance to RAS(ON) inhibitors, and identifies STAT3 inhibition as a rational combination strategy. These findings offer preclinical evidence supporting mutation-guided patient stratification and combination-therapy development for NRAS-mutated melanoma. Statement of Translational Significance NRAS-mutated melanoma lacks effective targeted treatments, and clinical responses to immunotherapy remain suboptimal. This study presents the first comprehensive drug sensitivity map across 95 NRAS mutations in melanoma, identifying the pan-RAS(ON) inhibitors RMC-6236 and RMC-7977 as broadly effective agents. Multiple mutants with reduced susceptibility are identified, providing mutation-specific guidance for patient selection and clinical trial stratification. Mechanistic analyses reveal that STAT3 activation functions as a key survival pathway under RAS(ON) blockade, and its inhibition markedly enhances the efficacy of pan-RAS(ON) inhibitors. These findings support mutation-guided utilization of RAS(ON) inhibition and highlight STAT3 co-targeting as a rational strategy to strengthen and prolong therapeutic responses in NRAS-mutant melanoma. Highlights - Generation of mutation-resolved functional and drug-sensitivity atlas for 95 NRAS missense variants covering >99% of all NRAS recurrent mutation - RMC-6236 and RMC-7977 shows broad and potent activity across major NRAS hotspot mutations - Defines NRAS mutations with reduced susceptibility to provide actionable information for clinical trial stratification - RAS(ON) inhibition and napabucasin shows synergistic anti-tumor activity in NRAS-mutated melanoma