Introduction: Cutaneous and uveal melanomas present distinct molecular characteristics and therapeutic responses, complicating their clinical management. Artificial intelligence (AI) is increasingly being applied to improve gene therapy and precision oncology strategies, offering promising solutions to overcome treatment resistance and tumor heterogeneity. Methods: This work examines peer-reviewed studies from the past five years, focusing on AI integration in melanoma gene therapy, multi-omics analysis, predictive modeling, and immunotherapy response prediction. Results: AI has shown potential to improve treatment planning by predicting therapeutic response and progression-free survival through radiomics-based models. A novel disulfidptosis-linked signature enhances survival prediction and treatment sensitivity, while HLA-DQA1 has emerged as a promising therapeutic target. Integration of single-cell and bulk RNA sequencing has uncovered key metastasis-related genes, and MethylMix analysis has identified methylation-altered genes affecting expression. In uveal melanoma, AI-driven data fusion linked 48 genes to metastasis, while hdWGCNA enabled mapping of gene modules to specific cell types for refined risk prediction. Machine learning models further optimize prognostic assessments and therapeutic response prediction. A composite decision index incorporating programmed cell death modes improves prognostic accuracy. The immune response score predicts overall survival in cutaneous melanoma, and the AI-derived Stem.Sig signature links cancer stemness to immunotherapy resistance. Deep learning has also been used to analyze TERT promoter mutations for metastatic potential, while a six-gene panel predicts anti-PD-1 therapy response. CSIRG-based models improve patient stratification, and logistic regression reaffirms AI’s prognostic value. Conclusions: Despite the growing success of AI in integrating multi-omics data, enhancing survival prediction, and guiding personalized melanoma therapy, challenges remain. These include limited high-quality, melanoma-specific datasets and algorithmic bias, which may compromise clinical reliability. Overcoming these barriers will require robust data governance, ethical AI model development, and stronger institutional policies.

Pancreatic cancer is among the most lethal malignancies, largely due to its asymptomatic progression and the absence of effective early detection methods. Urinary extracellular vesicles (EVs) represent a promising, non-invasive source of disease-specific biomarkers. Data-Independent Acquisition (DIA) mass spectrometry enables comprehensive proteomic profiling; however, independent analyses of precursor (MS1) and fragment (MS2) ion spectra often yield inconsistent results due to differing sources of signal interference and variability. Urine samples from five pancreatic cancer patients and five healthy controls were processed to enrich EVs via ultracentrifugation. DIA-MS was performed using an Orbitrap Eclipse MS with raw data analyzed using Spectronaut 19.1. MS1 and MS2 intensities were normalized to exosome protein markers. We developed a unified linear mixed-effects model (LMM) to concurrently analyze normalized MS1 and MS2 intensities, treating the MS1 and MS2 signals as technical replicates from the same biological sample. This model estimates protein abundance differences between groups while accounting for intra-group variability. Its performance was compared with Spectronaut quantitation by benchmarking against standard MS1 or MS2 methods using both simulated data and pancreatic cancer clinical data. The unified LMM consistently outperformed single-stream analyses, identifying a balanced and accurate set of differentially abundant proteins while avoiding overestimation. Simulations confirmed the model’s robustness, achieving higher true positive rates and lower false positive rates across varying sample sizes. The model also provided conservative estimates when MS1 and MS2 data diverged. While the LMM tends to be more powerful, Spectronaut’s methods are comparatively more conservative. In clinical data, the LMM further improved pathway enrichment outcomes, offering deeper biological insights. This LMM approach enhances the reliability of differential protein analysis in DIA-based proteomics by integrating MS1 and MS2 data. Applied to urinary EVs, it enables robust biomarker discovery for pancreatic cancer detection. This method holds significant promise for advancing personalized medicine through precise, non-invasive diagnostics.

Introduction

Pharmacogenomic (PGx)-guided prescribing can reduce adverse drug reactions and improve efficacy, yet clinical uptake is limited. A key barrier is reliance on complicated, text-dense reports, which are incompatible with routine clinical workflows. Large language models (LLMs) show promise in generating concise, context-specific recommendations, but prior work has largely focused on concept-based question-answering rather than realistic case interpretation. In this study, we examine LLMs' efficacy in interpreting PGx reports.

Methods

GPT-4.5 was deployed in Stanford’s HIPAA-compliant SecureGPT environment and paired with retrieval-augmented generation to inject CPIC guidance at inference. To create the test dataset, we created a set of synthetic PGx laboratory reports and accompanying medical histories. These cases covered a range of genotypes and clinical contexts. An expert human evaluator interpreted each case and authored a gold-standard consult note. The LLM generated consult notes that were compared against gold-standard notes quantitatively using ROUGE-L and BERTScore. The qualitative evaluation (LLM-as-a-judge and human evaluators) used a 5-point Likert scale across five quality domains (accuracy, clinical relevance, bias, risk-management, and hallucination).

Results:

Human expert ratings achieved overall scores of 0.91 ± 0.16, while the LLM-as-a-judge scoring produced 0.708 ± 0.17. Semantic similarity metrics showed a BERTScore precision of 0.822 ± 0.012, recall of 0.788 ± 0.018, and F1 of 0.805 ± 0.013. The direct lexical overlap was lower, with a ROUGE-L precision of 0.207 ± 0.084 and a recall of 0.270 ± 0.122.

Conclusion:

LLMs can achieve quality PGx outputs compared to those from human experts. However, our findings show variable performance in the metrics. NLP-based evaluations miss nuanced clinical data and lack flexibility in interpretation. This also raises concerns about the efficacy of NLP metrics for reliably evaluating high-impact clinical information and underscores the necessity of human evaluation. Key limitations included an inability to factor in phenoconversion and decisively synthesise information for drugs influenced by multiple pharmacogenes. Further validation with real-world patient data is in progress.

Abstract
Introduction:
Non-communicable diseases (such as hypertension, chronic obstructive pulmonary disease, and cardiovascular disease) represent a major cause of morbidity and mortality worldwide. Preventive medicine is the primary care approach to early detection and treatment. Preventive medicine aims to coordinate patient care and interventions based on patient-specific criteria to improve health outcomes and resource utilization. The purpose of this study is to evaluate the feasibility and short-term outcomes of personalized cancer screening in a primary care clinic.

Methods:
A prospective observational study was conducted in a primary care clinic over a three-month period. Patients aged 18–70 years were assessed using a structured assessment tool that included demographic data, current history, and the Meyer's risk profile (blood pressure, body mass index, fasting blood sugar, and advanced risk profile). The average risk category (low, high, and low) was determined by the patient's individualized consultations, personalized lifestyle recommendations, and personalized lessons.

Results:
Of the 150 patients, 38% were classified as high-risk, 42% as moderate-risk, and 20% as very high-risk. After three months of follow-up, 45% of patients with the disease showed improved blood pressure control, and 30% showed improved adherence to normal activity. The high-risk group also showed significant improvements in dietary and weight management.

Conclusion:
With a small number of people in healthcare, it should be feasible and effective to implement a patient-centered approach to health behaviors. This could play a significant role in reducing the long-term burden of involuntary treatment and patient-centered care in similar healthcare settings.

Background

Gastric cancer (GC) is one of the most prevalent malignant tumors worldwide, with incidence and mortality rates expected to rise in the coming years. Standard treatments typically involve surgical resection combined with chemotherapy using platinum-based agents, 5-fluorouracil (5-FU), or capecitabine. The discovery of novel biomarkers associated with GC progression is critical to improving therapeutic strategies and enhancing patient outcomes. In this study, we investigated the role of Hormonally Upregulated Neu tumor-associated Kinase (HUNK) in driving the progression of gastric cancer.

Methods

To evaluate the functional effects of HUNK overexpression, we conducted a cell viability assay in gastric cancer cell lines stably transfected with HUNK. To confirm the physical interaction between HUNK and p38 MAP kinase, we performed a co-immunoprecipitation assay. Additionally, to investigate the role of HUNK in apoptosis, we employed siRNA-mediated HUNK knockdown and assessed apoptotic activation by measuring the levels of cleaved PARP and cleaved caspase-3 via Western blot. Apoptotic cell populations were further quantified using flow cytometry after staining with Annexin V-FITC and propidium iodide (PI). NGS analysis was performed to identify HUNK-induced genes.

Results

We demonstrate that HUNK promotes gastric cancer cell proliferation through direct interaction and phosphorylation of p38 MAP kinase. Functional depletion of HUNK significantly reduced cell viability in both gastric cancer cell lines and patient-derived organoids, underscoring its role in tumor cell survival. Moreover, HUNK positively regulates the expression of MUC16/CA-125, a glycoprotein linked to tumor progression and poor prognosis, suggesting its potential as both a biomarker and therapeutic target in gastric cancer.

Conclusions

Our results identify a novel HUNK-mediated molecular mechanism in gastric cancer, supporting its potential as a promising target for the development of targeted therapies.

Introduction:
Advancements in transcriptomic profiling and machine learning are transforming preventive oncology. Triple-Negative Breast Cancer (TNBC), a clinically aggressive and heterogeneous subtype, lacks targeted therapies and presents challenges for early interception. This study investigates the potential of unsupervised learning to reveal molecular subtypes within TNBC, supporting individualized risk assessment strategies.

Methods:
Transcriptomic and clinical data from the TCGA breast cancer cohort were analyzed. TNBC cases were identified based on basal-like classification. Dimensionality reduction was performed using Principal Component Analysis (PCA), followed by k-means clustering to detect latent subgroups. Genes with the highest variance were used to explore inter-sample expression patterns.

Results:
The PCA of TNBC transcriptomic data revealed distinct variance among patient samples, indicating underlying molecular heterogeneity. The first two principal components captured a substantial portion of the variance, allowing for clear visual separation. K-means clustering (k=3) identified three reproducible molecular subgroups with minimal overlap and strong intra-cluster similarity. Analysis of the top 50 most variably expressed genes showed distinct expression patterns across clusters, suggesting differential pathway activation. Several of these genes are linked to cancer progression, immune response, and therapy resistance, highlighting their potential as early biomarkers. These findings demonstrate that unsupervised transcriptomic clustering can uncover clinically relevant TNBC subtypes, supporting subtype-specific risk stratification and preventive strategies.

Conclusions:
This study presents a reproducible framework for uncovering transcriptional heterogeneity in TNBC using unsupervised machine learning. By enabling early subgroup identification, this approach supports the goals of personalized preventive medicine through data-driven stratification and future biomarker discovery.

Inflammatory Bowel Disease (IBD), including Crohn’s disease and ulcerative colitis, presents a major clinical challenge due to its chronic nature and the limitations of existing therapies. In this context, patient-derived organoids (PDOs) have emerged as powerful preclinical models that preserve the cellular complexity, architecture, and molecular features of the original intestinal tissues, making them ideal tools for both disease modeling and drug screening.

In this study, we established intestinal PDOs from both healthy individuals and IBD patients, demonstrating their ability to recapitulate key features of the native tissue, including the expression of lineage-specific markers and pro-inflammatory mediators. Notably, the IBD-derived organoids retained elevated levels of IL-8 and CLDN1, reflecting their inflammatory state and validating them as a faithful in vitro model of the disease.

Leveraging this platform, we investigated the anti-inflammatory potential of an ozonated extra virgin olive oil derivative, using a luciferase reporter system driven by NF-κB response elements. Upon TNFα-induced inflammation, this natural compound significantly reduced the NF-κB activation in both 2D cell lines and 3D organoid systems, showing comparable efficacy to that of standard anti-inflammatory compounds such as dexamethasone. These findings were further supported by immunofluorescence, which revealed decreased nuclear translocation of p65 following treatment.

Importantly, this compound displayed minimal cytotoxicity at effective concentrations in preliminary viability assays.

Overall, this study underscores the critical value of PDOs in faithfully modeling IBD and evaluating novel therapeutic agents. Our results highlight an olive oil derivative as a promising natural compound with strong anti-inflammatory effects and pave the way for PDO-based personalized medicine approaches in chronic intestinal inflammation. This organoid-driven strategy offers a robust and translational platform for screening natural products and developing safer, more targeted treatments for IBD.

Despite antipsychotic drugs being a key treatment for schizophrenia, 20–30% of patients show an inadequate response. This may be due to genetic factors affecting drugs' metabolism, side effects, and efficacy. This study explored links between genetic polymorphisms and pharmacoresistant (PR) schizophrenia in Belarusian patients.

A total of 161 participants were included: 104 with PR schizophrenia (no response after 6–8 weeks of treatment) and 57 responders. All were diagnosed according to the ICD-10 (F20), aged 18–60, gave informed consent, and had no acute physical illnesses. The treatments included clozapine (42.1%) or combined atypical and typical antipsychotics (39.5%). We genotyped 19 polymorphic loci across 15 genes (e.g., CYP2D6, COMT, HTR1A, MDR1) using a Real-Time PCR or Sanger sequencing. The data were analyzed in SPSS 20.0 using ORs and 95% CIs to evaluate the genotype–phenotype associations.

Significant associations with PR were found for the following:

• The A-allele and AG genotype of CYP2D6*4 (rs3892097) (OR=2.159 and CI=1.018-4.578 and OR=2.975 and CI=1.269-6.977, respectively);
• The G-allele of COMT (rs4680, V158) (OR=2.786; CI=1.414-5.489);
• The CC genotype of HTR1A (rs6295) (OR=2.703; CI=1.151-6.348).

Protective effects were noted for the following:

• The GG genotype of CYP2D6 (rs3892097) (OR=0.463; CI=0.218-0.928);
• The G allele of HTR1A (rs6295) (OR=0.367; CI=0.156-0.864);
• The AA genotype (Met/Met) of COMT (rs4680) (OR=0.359; CI=0.182-0.707).

Combined genotypes showed an even greater PR risk, including the following:

• A-/A- (CYP2D6/CYP1A2) (OR = 2.926; 95% CI = 1.206–7.102);
• A-/A-/T- (CYP2D6/CYP1A2/MDR1) (OR = 4.833; 95% CI = 1.753–13.328)
• G-/LL (COMT/SLC6A4) (OR = 6.923; 95% CI = 1.900–25.227);
• CC/T- (HTR1A/MDR1) (OR = 2.564; 95% CI = 1.120–5.873).

In conclusion, variants in CYP2D6, HTR1A, and COMT, especially when combined with variants in CYP1A2, MDR1, and SLC6A4, significantly contribute to the PR schizophrenia risk in Belarus. These findings support the use of pharmacogenetic testing for personalized antipsychotic treatment, already implemented in the Republican Research and Practice Center for Mental Health (Minsk).

Abstract: this study aims to gather dopamine (PDA) modified engineering bacteria load glucose oxidase (bigger) and former drug taxol, and realize the tumor targeting hungry treatment and the synergy of taxol drug delivery. By taking advantage of the specific tumor-targeting characteristics of Escherichia coli as a carrier, the recognition and endocytosis ability of tumor cells are enhanced through the camouflage properties of its cell membrane. GOx catalyzes the oxidation of glucose within tumor cells to produce hydrogen peroxide (HO₂), thereby inducing a "starvation" state in tumor cells and inhibiting their proliferation and metabolism. Meanwhile, paclitaxel prodrugs are reduced to active paclitaxel in the tumor microenvironment (TME), exerting anti-tumor effects. This system achieves the synergistic treatment of tumors through a dual mechanism.

The results indicate that PDA-modified E. coli co-loaded with GOx and a paclitaxel prodrug exhibits considerable inhibitory effects against various tumor cell lines in vitro, as well as effective tumor suppression in vivo. Compared with the administration of either GOx or paclitaxel alone, this integrated therapeutic system achieves enhanced treatment outcomes in terms of both tumor growth inhibition and survival extension. Moreover, the system demonstrates favorable site-specific drug release behavior, which effectively mitigates tumor hypoxia and consequently improves overall therapeutic performance.

Cytochrome P450 3A4 (CYP3A4) and P-glycoprotein (P-gp) are critical regulators of drug metabolism and disposition, with significant implications for inter-individual variability in pharmacokinetics and therapeutic outcomes. CYP3A4 is a major hepatic enzyme responsible for the oxidative biotransformation of a wide range of drugs, while P-gp functions as an efflux transporter, actively exporting drugs out of cells and limiting intracellular drug accumulation. Notably, previous studies have demonstrated a functional interplay between CYP3A4 and P-gp, particularly in tissues where they are co-expressed, such as the intestinal epithelium. Alterations in P-gp and CYP3A4 activity—whether due to genetic variation or to co-administered inhibitors or inducers—can significantly affect drug pharmacokinetics (absorption, distribution, metabolism, and excretion—ADME), posing challenges in dose optimization and drug safety.

In this study, we investigated the structural–functional crosstalk between CYP3A4 and P-gp through protein–protein interaction (PPI) modeling using three different tools: Rosetta protein–protein docking, the Schrödinger pipeline for protein–protein docking, and AlphaFold 3. The resulting complexes were refined and validated via molecular dynamics simulations. We further examined the impact of CYP3A4–P-gp interactions on the metabolism and transport of three Janus kinase (JAK) inhibitors—tofacitinib, baricitinib, and ruxolitinib—which are widely used in autoimmune disease therapy. Molecular docking and simulation analyses revealed how the CYP3A4–P-gp interaction may influence the binding behavior and metabolic handling of these drugs. Preliminary results suggest that physical association between the two proteins may alter substrate accessibility and retention, potentially modulating drug bioavailability and the drug clearance profile. Finally, we constructed an in silico pharmacokinetic compartment model to simulate drug disposition under varying enzymatic and transporter activity.

Our findings highlight the importance of integrating PPI, molecular pharmacology, and pharmacokinetics modeling to advance personalized medicine approaches in immunology.