Gastrointestinal (GI) cancers constitute an umbrella term for a wide variety of malignancies that are located in the digestive tract (esophageal, gastric, small and large intestine, anus, liver, gallbladder, and pancreas), with 25% of total cancers and 35% of cancer-related deaths being attributed to them. An alarming trend of rising GI malignancy diagnoses, especially in younger age groups, underscores the need for discoveries in liquid-based biomarkers that facilitate both early detection and optimal disease management. Extracellular vesicles (EVs) not only constitute promising nano-sized biomarkers, but also, via bioengineering, have shown a great therapeutic potential, with artificial intelligence (AI) revolutionizing their research via the selection of the best biomarkers from omics, the recognition of pathophysiological patterns, and facilitating a faster drug-development via AI-driven EV engineering, drug delivery modeling, and target identification. In this review, we will provide a clear insight into the implementation of AI methodologies in EV-based biomarker discovery and therapeutics for pancreatic and hepatobiliary cancer.

As treatment plans involve more medications, anticipating and preventing drug-drug interactions (DDIs) becomes increasingly important. Such interactions can result in harmful side effects and may reduce therapy effectiveness. Currently, most computational approaches for DDI prediction rely heavily on complex feature engineering and require chemical information to be structured in specific formats to enable accurate detection of potential interactions. This study presents the first investigation of the application of Large Language Models (LLMs) for DDI prediction using drug characteristics expressed solely in free-text form. Specifically, we use SMILES notations, target organisms, and gene associations as inputs in purpose-designed prompts, allowing LLMs to learn the underlying relationships among these descriptors and accordingly predict possible DDIs. We evaluated the performance of 18 distinct LLMs under zero-shot, few-shot, and fine-tuning settings on the DrugBank dataset (version 5.1.12) to identify the most effective paradigm. We then assessed the generalizability of the fine-tuned models on 13 external DDI datasets against well-known machine learning baselines. The results demonstrated that, while zero-shot and few-shot paradigms showed only modest utility, fine-tuned models achieved superior sensitivity while maintaining competitive accuracy and F1-score compared to baselines. Notably, despite its small size, the Phi-3.5 2.7B model attained a sensitivity of 0.978 and an accuracy of 0.919. These findings suggest that computational efficiency and task-specific adaptation are more important than model size in order to capture the complex patterns inherent in drug interactions, and outline a more accessible paradigm for DDI prediction that can be integrated into clinical decision support systems.

Advanced fibrosis is a major determinant of liver-related morbidity in metabolic dysfunction-associated steatotic liver disease (MASLD). FIB-4 is widely used as a first-line non-invasive test, but its fixed formula may underuse diagnostic information contained in age, aspartate aminotransferase, alanine aminotransferase, and platelet count. We evaluated whether machine-learning-enhanced non-invasive testing (MLE-NIT) can improve advanced fibrosis detection while preserving this FIB-4 variable space. We used three biopsy-confirmed MASLD cohorts from China, Malaysia, and India (n=784). The Chinese cohort was split into 486 training and 54 internal validation/tuning patients; final performance was reported only on the Malaysian and Indian external cohorts. Models used five variables: age, FIB-4, aspartate aminotransferase, platelet count, and alanine aminotransferase. We compared FIB-4 with a shallow-deep neural network (s-DNN), TabPFN, and gpt-4o-2024-08-06. FIB-4 achieved external ROC-AUCs of 0.75 and 0.60 in Malaysia and India, respectively. TabPFN achieved 0.69 and 0.66, fine-tuned GPT-4o achieved 0.75 and 0.63, and the s-DNN achieved 0.77 and 0.67, respectively. The s-DNN contained only 354 trainable parameters, compared with 7,244,554 for TabPFN, yet provided a more balanced external operating profile. Calibration showed s-DNN Brier scores of 0.18 and 0.22, and permutation importance identified AST and FIB-4 as dominant variables. Compact non-linear MLE-NITs may enhance FIB-4-based fibrosis assessment without increasing clinical data requirements.

BACKGROUND:Metabolic dysfunction-associated steatotic liver disease (MASLD) is a leading cause of chronic liver disease globally. Current diagnostic methods, such as liver biopsies, are invasive and have limitations, highlighting the need for non-invasive alternatives. AIM:To investigate extracellular vesicles (EVs) as potential biomarkers for diagnosing and staging steatosis in patients with MASLD using machine learning (ML) and explainable artificial intelligence (XAI). METHODS:In this single-center observational study, 798 patients with metabolic dysfunction were enrolled. Of these, 194 met the eligibility criteria, and 76 successfully completed all study procedures. Transient elastography was used for steatosis and fibrosis staging, and circulating plasma EV characteristics were analyzed through nanoparticle tracking. Twenty ML models were developed: Six to differentiate non-steatosis (S0) from steatosis (S1-S3); and fourteen to identify severe steatosis (S3). Models utilized EV features (size and concentration), clinical (advanced fibrosis and presence of type 2 diabetes mellitus), and anthropomorphic (sex, age, height, weight, body mass index) data. Their performance was assessed using receiver operating characteristic (ROC)-area under the curve (AUC), specificity, and sensitivity, while correlation and XAI analysis were also conducted. RESULTS:The CatBoost C1a model achieved an ROC-AUC of 0.71/0.86 (train/test) on average across ten random five-fold cross-validations, using EV features alone to distinguish S0 from S1-S3. The CatBoost C2h-21 model achieved an ROC-AUC of 0.81/1.00 (train/test) on average across ten random three-fold cross-validations, using engineered features including EVs, clinical features like diabetes and advanced fibrosis, and anthropomorphic data like body mass index and weight for identifying severe steatosis (S3). Key predictors included EV mean size and concentration. Correlation, XAI, and SHapley Additive exPlanations analysis revealed non-linear feature relationships with steatosis stages. CONCLUSION:The EV-based ML models demonstrated that the mean size and concentration of circulating plasma EVs constituted key predictors for distinguishing the absence of significant steatosis (S0) in patients with metabolic dysfunction, while the combination of EV, clinical, and anthropomorphic features improved the diagnostic accuracy for the identification of severe steatosis. The algorithmic approach using ML and XAI captured non-linear patterns between disease features and provided interpretable MASLD staging insights. However, further large multicenter studies, comparisons, and validation with histopathology and advanced imaging methods are needed.

Medication errors significantly threaten patient safety, leading to adverse drug events and substantial economic burdens on healthcare systems. Clinical Decision Support Systems (CDSSs) aimed at mitigating these errors often face limitations when processing unstructured clinical data, including reliance on static databases and rule-based algorithms, frequently generating excessive alerts that lead to alert fatigue among healthcare providers. This paper introduces HELIOT, an innovative CDSS for adverse drug reaction management that processes free-text clinical information using Large Language Models (LLMs) integrated with a comprehensive pharmaceutical data repository. HELIOT leverages advanced natural language processing capabilities to interpret medical narratives, extract relevant drug reaction information from unstructured clinical notes, and learn from past patient-specific medication tolerances to reduce false alerts, enabling more nuanced and contextual adverse drug event warnings across primary care, specialist consultations, and hospital settings. Evaluation using three state-of-the-art LLMs on synthetic and real-world datasets demonstrates classification accuracy ranging from 98.77% to 99.80% with zero false negatives for life-threatening reactions. This high accuracy enabled HELIOT to achieve a 50-53 % reduction in interruptive alerts compared to traditional CDSSs while maintaining perfect safety profiles. To support clinical deployment, the system incorporates a confidence-based risk stratification framework that enables automated decisions for high-certainty cases while ensuring appropriate clinical oversight for uncertain classifications. Clinical usability evaluation with healthcare professionals validated these achievements, revealing strong acceptance and unanimous preference for HELIOT's contextual approach over traditional systems. These findings show promise; however, broader clinical trials remain essential to confirm effectiveness across diverse healthcare environments.

Background & Aims: Metabolic dysfunction-associated steatotic liver disease (MASLD), formerly known as NAFLD, is a leading cause of chronic liver disease worldwide. Current diagnostic methods, including liver biopsies, are invasive and have significant limitations, emphasizing the need for non-invasive alternatives. This study aimed to evaluate extracellular vesicles (EV) as biomarkers for diagnosing and staging steatosis in MASLD patients, utilizing machine learning (ML) and explainable artificial intelligence (XAI). Methods: This prospective, single-center cohort study was conducted at the GI-Liver Unit, Hippocration General Hospital, Athens. It included 76 MASLD patients with ultrasound-confirmed steatosis and at least one cardiometabolic risk factor. Patients underwent transient elastography for steatosis and fibrosis staging and blood sampling for EV analysis using nanoparticle tracking. Twenty machine learning models were developed. Six to distinguish non-steatosis (S0) from steatosis (S1-S3), and fourteen to identify severe steatosis (S3). Models incorporated EV measurements (size and concentration), anthropomorphic and clinical features, with performance evaluated using AUROC and SHAP-based interpretability methods. Results: The CB-C1a model achieved, on average on 10 random splits of 5-fold cross validation (5CV) of the train set, an AUROC of 0.71/0.86 (train/test) for distinguishing S0 from S1-S3 steatosis stages, relying on EV alone. The CB-C2h-21 model identified severe steatosis (S3), on average on 10 random splits of 3-fold cross validation (3CV) of the train set, with an AUROC of 0.81/1.00 (train/test), demonstrating superior performance when combining EV with anthropomorphic and clinical features such as diabetes and advanced fibrosis. Key EV features, including mean size and concentration, were identified as important predictors. SHAP analysis highlighted complex non-linear relationships between features and steatosis staging. Conclusions: EV are promising non-invasive biomarkers for diagnosing and staging MASLD. The integration of ML-enhanced EV analysis with clinical features offers a scalable, patient-friendly alternative to invasive liver biopsies, advancing precision in MASLD management. Further research should refine these methods for broader clinical application. Keywords: Metabolic dysfunction-associated steatotic liver disease (MASLD), extracellular vesicles (EVs), non-invasive biomarkers, machine learning (ML), explainable artificial intelligence (XAI), steatosis staging, transient elastography, chronic liver disease, cardiometabolic risk factors, SHAP analysis, diagnostic precision, advanced fibrosis, severe steatosis (S3), liver biopsy alternatives, hepatology diagnostics. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement This study did not receive any funding ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: Ethics Committee of General Hospital of Athens "Hippocratio" in 1st Health Authority of Greece ( Attica) gave ethical approval for this work I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes All data produced in the present study are available upon reasonable request to the authors

BackgroundDespite the impressive outcomes with immune checkpoint inhibitor (ICI) in non-small cell lung cancer (NSCLC), only a minority of the patients show long-term benefits from ICI. In this study, we used retrospective cohorts of ICI treated patients with NSCLC to discover and validate spatially resolved protein markers associated with resistance to programmed cell death protein-1 (PD-1) axis inhibition.MethodsPretreatment samples from 56 patients with NSCLC treated with ICI were collected and analyzed in a tissue microarray (TMA) format in including four different tumor regions per patient using the GeoMx platform for spatially informed transcriptomics. 34 patients had assessable tissue with tumor compartment in all 4 TMA spots, 22 with leukocyte compartment and 12 with CD68 compartment. The patients’ tissue that was not assessable in fourfold redundancy in each compartment was designated as the validation cohort; cytokeratin (CK) (N=22), leukocytes CD45 (N=31), macrophages, CD68 (N=43). The human whole transcriptome, represented by~18,000 individual genes assessed by oligonucleotide-tagged in situ hybridization, was sequenced on the NovaSeq platform to quantify the RNAs present in each region of interest.Results54,000 gene variables were generated per case, from them 25,740 were analyzed after removing targets with expression lower than a prespecified frequency. Cox proportional-hazards model analysis was performed for overall and progression-free survival (OS, PFS, respectively). After identifying genes significantly associated with limited survival benefit (HR>1)/progression per spot per patient, we used the intersection of them across the four TMA spots per patient. This resulted in a list of 12 genes in the tumor-cell compartment (RPL13A, GNL3, FAM83A, CYBA, ACSL4, SLC25A6, EPAS1, RPL5, APOL1, HSPD1, RPS4Y1, ADI1). RPL13A, GNL3in tumor-cell compartment were also significantly associated with OS and PFS, respectively, in the validation cohort (CK: HR, 2.48; p=0.02 and HR, 5.33; p=0.04). In CD45 compartment, secreted frizzled-related protein 2, was associated with OS in the discovery cohort but not in the validation cohort. Similarly, in the CD68 compartmentARHGAPand PNN interacting serine and arginine rich protein were significantly associated with PFS and OS, respectively, in the majority but not all four spots per patient.ConclusionThis work highlightsRPL13AandGNL3as potential indicative biomarkers of resistance to PD-1 axis blockade that might help to improve precision immunotherapy strategies for lung cancer.

Data augmentation (DA) enhances model generalization in computer vision but may introduce biases, impacting class accuracy unevenly. Our study extends this inquiry, examining DA's class-specific bias across various datasets, including those distinct from ImageNet, through random cropping. We evaluated this phenomenon with ResNet50, EfficientNetV2S, and SWIN ViT, discovering that while residual models showed similar bias effects, Vision Transformers exhibited greater robustness or altered dynamics. This suggests a nuanced approach to model selection, emphasizing bias mitigation. We also refined a "data augmentation robustness scouting" method to manage DA-induced biases more efficiently, reducing computational demands significantly (training 112 models instead of 1860; a reduction of factor 16.2) while still capturing essential bias trends.

Applying the iterative methodology for dimensionality reduction/feature selection using categorical gradient boosted trees, as it has been defined in and has been successfully applied on similar datasets in and , on a dataset consisted of 12708 gene expressions coming from 5052 individuals from 105 studies, we classify whether a person has acute myeloid leukaemia (AML) or is healthy. A CatBoost model on a dataset with reduced dimensions of 72 genes reached a ROC-AUC score of 0.9973 using ten fold cross validation (10CV) and ROC-AUC: 0.9988 on an inference dataset. We further investigate the options of using less genes that potentially could be used in clinical practise and genes than have not been associated to AML yet, or to blood cancer in general. On the same folds of the 10CV and on the same inference dataset the performance of the tuned CatBoost models show that it could be the case that not all genes with an association to AML have been found yet and 19 genes could be enough to predict AML: CatBoost63 (ROC-AUC: 0.9941, Test: 0.9942), CatBoost19: (ROC-AUC: 0.9946, Test: 0.9941) and CatBoost15: (ROC-AUC: 0.9922, Test: 0.9900). In addition, our results verify that a gene diagnostic test for AML could be possible in the future as well as further research is needed on these 15 genes that it could lead to new and better drugs.