Find the papers that actually matter

Survival prognosis is crucial for medical informatics. Practitioners often confront small-sized clinical data, especially cancer patient cases, which can be insufficient to induce useful patterns for survival predictions. This study deals with small sample survival analysis by leveraging transfer learning, a useful machine learning technique that can enhance the target analysis with related knowledge pre-learned from other data. We propose and develop various transfer learning methods designed for common survival models. For parametric models such as DeepSurv, Cox-CC (Cox-based neural networks), and DeepHit (end-to-end deep learning model), we apply standard transfer learning techniques like pretraining and fine-tuning. For non-parametric models such as Random Survival Forest, we propose a new transfer survival forest (TSF) model that transfers tree structures from source tasks and fine-tunes them with target data. We evaluated the transfer learning methods on colorectal cancer (CRC) prognosis. The source data are 27,379 SEER CRC stage I patients, and the target data are 728 CRC stage I patients from the West China Hospital. When enhanced by transfer learning, Cox-CC's Ctd value was boosted from 0.7868 to 0.8111, DeepHit's from 0.8085 to 0.8135, DeepSurv's from 0.7722 to 0.8043, and RSF's from 0.7940 to 0.8297 (the highest performance). All models trained with data as small as 50 demonstrated even more significant improvement. Conclusions: Therefore, the current survival models used for cancer prognosis can be enhanced and improved by properly designed transfer learning techniques. The source code used in this study is available at https://github.com/YonghaoZhao722/TSF.

Tumor-infiltrating lymphocytes (TILs) are an important indicator of immune activity in breast cancer, yet scoring them consistently on H&E slides remains challenging in routine pathology. This work presents a modular deep learning pipeline that delivers fully automated and continuous stromal TILs (sTILs) scores in line with the International Immuno-Oncology Biomarker Working Group (IIOBWG) guidelines. The pipeline combines three components: a TIL segmentation model refined through pathologist-guided active learning, a robust stroma segmentation network based on an enhanced DeepLabV3+, and a lightweight regression module that learns how TILs distribute within stromal regions. A new adaptive aggregation strategy integrates patch-level predictions into a single, clinically meaningful score while accounting for heterogeneous infiltration. The system was evaluated on two independent datasets (60 and 112 WSIs) with expert-annotated ROIs, achieving strong agreement with pathologists (Pearson of 0.814; ICC of 0.808). Importantly, the pipeline is interpretable: each stage produces human-readable outputs (stroma masks, TIL-in-stroma maps), and SegGradCAM visualizations confirm that predictions rely on biologically relevant tissue regions. These findings demonstrate the pipeline's potential as a reliable and clinically adaptable tool for standardized, fully automated TILs quantification in breast cancer pathology. The source code and pretrained models are publicly available at https://github.com/Shrief-Abdelazeez/TILs-Scoring.

Liver cancer, particularly hepatocellular carcinoma (HCC), is a significant global health concern due to its asymptomatic early stages, biological diversity, and frequent late diagnoses that hinder effective treatment and survival rates. Traditional diagnostic methods, such as serum biomarker assays and imaging techniques, often lack the necessary sensitivity and specificity and highlight the urgent need for innovative, non-invasive diagnostic alternatives. This review emphasizes the potential of combining nanobiosensor technologies with physics-informed machine learning (PIML) to address these diagnostic challenges. Nanobiosensors utilize advanced materials like gold nanoparticles and graphene to achieve highly sensitive, real-time detection of HCC biomarkers, including alpha-fetoprotein (AFP) and non-coding RNAs, with detection limits reaching sub-nanomolar to femtomolar levels through various mechanisms. However, the clinical application of nanobiosensors is hindered by issues such as signal instability and environmental interference. PIML offers a solution by incorporating fundamental physical principles into machine learning models which is enhancing their predictive accuracy and robustness against data noise. This hybrid approach facilitates effective signal denoising, adaptive calibration, and the integration of multimodal data, thereby improving the overall diagnostic process. Main findings indicate that PIML-enhanced nanobiosensing systems significantly outperform traditional AI models in biomedical applications, demonstrating superior generalization and biologically relevant outputs even in the presence of limited data. The integration of these technologies creates a promising framework for advanced liver cancer diagnostics, enabling precise, non-invasive detection and personalized clinical decision-making. In conclusion, the convergence of nanobiosensors and PIML holds the potential to revolutionize liver cancer diagnostics, offering improved early detection and dynamic monitoring. However, to realize this potential, ongoing challenges related to computational scalability, sensor reproducibility, and regulatory validation must be systematically addressed through collaborative interdisciplinary efforts.

Histone deacetylase 2 (HDAC2) plays a critical role in the pathogenesis of diffuse large B-cell lymphoma (DLBCL), positioning it as an attractive therapeutic target. In this study, we applied an integrated computational strategy that combined machine learning-based quantitative structure-activity relationship (QSAR) modelling, molecular docking, ADMET filtering, and molecular dynamics (MD) simulations to identify and optimize potential HDAC2 inhibitors. A dataset of 1995 HDAC2-active molecules was assembled and reduced to 25 key molecular descriptors, enabling the development of a robust Random Forest QSAR model (R2 = 0.926, CCC = 0.956). Top-predicted compounds underwent docking and pharmacokinetic evaluation, highlighting compound 10 as a promising lead (binding energy ≈ -10.2 kcal·mol-1). Guided optimization produced analogue 10 g, which displayed enhanced affinity (-10.9 kcal·mol-1), stable protein-ligand interactions in MD simulations, and favourable MM-GBSA binding free energy (ΔG_bind ≈ -40.3 kcal·mol-1). In silico pharmacokinetic analysis further suggested acceptable safety and drug-likeness. This multi-stage pipeline prioritizes 10 g as a strong candidate for experimental validation and demonstrates the value of integrating predictive modelling with structure-based refinement in anti-lymphoma drug discovery.

Treatment outcome prediction plays an important role in realizing personalized cancer therapy. In triple-negative breast cancer (TNBC), neoadjuvant chemotherapy (NAC) is widely used to downstage tumors and improve surgical outcomes. In head and neck cancer (HNC), early prediction of lesion progression can assist treatment planning. However, inter-patient heterogeneity in treatment response and tumor behavior limits the effectiveness of generalized treatment strategies. To address this issue, we developed an evidential reasoning rule-enabled deep neural network (ER2-DNN) for reliable outcome prediction in cancer therapy. The ER2-DNN combines convolutional neural network (CNN) based image feature extraction with data augmentation, Monte Carlo dropout, test-time augmentation and evidential reasoning rule (ER2) fusion for generating uncertainty-aware prediction. Across both TNBC and HNC datasets, the model showed consistent predictive performance with well-calibrated confidence estimates. The ER2-DNN provides a framework for supporting individualized oncology decisions through reliable image-based modeling.

Triple-negative breast cancer (TNBC) is an aggressive malignancy often characterized by chemoresistance, partly due to the overexpression of drug efflux transporters like ABCG2. To address this challenge, this study developed and evaluated a cRGD-modified, pH-sensitive liposomal system for the targeted co-delivery of docetaxel (DTX) and siRNA against ABCG2 (si-ABCG2). The synthesized nanoparticles (DTX/siRNA/cRGD-PLPs) exhibited optimal physicochemical properties, including a mean particle size of approximately 241.7 nm, efficient co-loading of DTX and siRNA, and pH-responsive cargo release, while protecting the siRNA from degradation in serum. Also, DTX/siRNA/cRGD-PLPs maintained homogeneous size distributions over the storage period and induced minimal hemoglobin release, with hemolysis rates remaining below safety threshold. These liposomes demonstrated enhanced, time-dependent uptake into TNBC cell lines HCC1937 and MDA-MB-231. In vitro, the DTX/siRNA/cRGD-PLPs formulation was significantly more effective at inhibiting cell viability, proliferation, migration, and invasion, and at inducing apoptosis, compared to the free drug combinations and other controls. Moreover, the dual-payload co-delivery liposomes (DTX/siRNA/cRGD-PLPs) exerted superior anti-tumor effects relative to single-agent formulations. In an MDA-MB-231 xenograft mouse model, the liposomal treatment was well-tolerated and resulted in marked tumor growth inhibition, which was associated with reduced cell proliferation (Ki67) and increased apoptosis (Caspase-3) within the tumor tissue. This targeted co-delivery system shows significant potential for improving TNBC treatment by synergistically enhancing therapeutic efficacy and overcoming chemoresistance.

Analyzing gene expression data is essential for predicting and detecting diseases, including cancer. The data is very repetitive and noisy, which makes it hard to find important information about illnesses. In the past decade, several traditional machine learning and feature selection models have been developed for cancer type classification from gene expression data. Rather than introducing new deep learning primitives, this work presents a principled integration framework that combines sparsity-aware representation learning, structure-inducing spatial embedding, and hierarchical multi-scale attention. This paper presents a method for cancer gene classification based on Hierarchical Attention Assisted Feature Pyramid Network (HA-FPN). The work involved two publicly available datasets. The proposed methodology starts with dimensionality reduction through a Variational Sparse Autoencoder (VSAE), followed by an updated DeepInsight algorithm for image conversion of the input. Next, the classification technique is constructed using the proposed HA-FPN model. Moreover, the improved gradient descent optimization (IGDO) is utilized to change the hyperparameter of the classification model. In addition, the results demonstrate that the model combined with an IGDO outperforms the currently existing methods in terms of accuracy, precision, recall, and F1-score. The method that is presented efficiently brings out different aspects of the data through t-SNE computations. Moreover, the proposed approach is very robust, and hence, it can reach high performance levels on two different datasets.

Neurotrophic tyrosine receptor kinase 2 (NTRK2/TRKB) demonstrates oncogenic roles across cancers, with notable significance in gliomas where its overexpression is linked to aggressive clinical phenotypes. Lucitanib (AL3810), a multi-target tyrosine kinase inhibitor, shows unexplored potential for treating NTRK2-driven gliomas. This study employed an integrated approach combining pan-cancer analysis, computational drug screening, and experimental validation to systematically evaluate the oncogenic function of NTRK2 and the therapeutic efficacy of Lucitanib. Multi-omics analysis across 32 cancer types using TCGA/GTEx/CPTAC datasets revealed NTRK2 overexpression in gliomas (GBMLGG/LGG), correlating with poor prognosis (p < 0.01) and implicating AKT signaling and immune microenvironment modulation. Structure-based virtual screening of 26,996 compounds against NTRK2 identified Lucitanib as a high-affinity binder (ΔG < -8 kcal/mol), forming stable interactions with key residues (MET-636, PHE-633), further validated by 500 ns molecular dynamics simulations. In vitro experiments using U251MG glioblastoma cells and primary astrocytes demonstrated that Lucitanib significantly inhibited proliferation (p < 0.001), suppressed invasion and migration via MMP9 downregulation (p < 0.001), and induced apoptosis through Bcl-2/Bax modulation (p < 0.001). Its efficacy was intermediate between methotrexate and the selective NTRK2 inhibitor ana-12. Mechanistically, Lucitanib targeted the NTRK2-AKT-MMP9 axis while preserving immune effector functions. These findings establish NTRK2 as a viable therapeutic target in gliomas and highlight Lucitanib as a novel multi-mechanistic inhibitor with balanced efficacy and favorable pharmacokinetic properties, supporting its further development for clinical translation in NTRK2-overexpressing gliomas.

Type 2 diabetes (T2D) places substantial physiological and psychological demands on patients and is independently linked to an elevated risk of depression. Intellectual developmental disability (IDD) is likewise associated with metabolic disorders and a high prevalence of mood disturbances, yet communication barriers often delay diagnosis. Whether coexistence of IDD further amplifies the likelihood of new-onset depression in people with T2D remains unclear. We aimed to investigate the association between IDD and incident depression among Korean adults with T2D. We analyzed 1,819,869 adults (≥ 20 years) with T2D who underwent the 2015-2016 Korea National Health Screening Program. Participants were classified as either IDD (n = 3665) or non-IDD groups. The primary outcome was new-onset depression identified up to 31 December 2022 following the health-screening date. Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for incident depression. Participants with IDD were younger (mean age: 49.2 vs. 58.0 years) and had a lower proportion of men (55.9% vs. 61.8%) than those without IDD. Over a median follow-up of 5.8 years, 14.8% developed depression, with an elevated risk in the IDD group (HR 1.65, 95% CI: 1.53-1.77). This association was consistent across IDD severity and was especially marked in individuals under 65 and with T2D duration under five years. Coexisting IDD and T2D are linked to higher depression risk. Our finding underscores the need for tailored interventions, improved caregiver awareness, and enhanced screening to address mental health disparities.

Chemodynamic therapy (CDT) based on overproduced reactive oxygen species (ROS), activates cytoprotective autophagy, an inexorable phenomenon-that enables tumor cell survival, therefore attenuating ROS-induced therapeutic efficacy. Herein, we develop a tumor microenvironment (TME)-responsive nanoplatform (MnOx-GOx-PM@Ma) composed of manganese oxide nanoflowers (MnOx NFs) co-loaded with glucose oxidase (GOx) and an activatable melittin pro-peptide (PM), and coated with macrophage membranes (Ma) for targeted delivery. Combined MnOx NFs and GOx trigger O2/H2O2 cyclic generation, thereby amplifying CDT in the acidic and glutathione (GSH)-rich TME. Meanwhile, the PM is selectively cleaved by lysosomal legumain to activate melittin, which disrupts lysosomal membranes and converts cytoprotective autophagy into a pro-death process. Additionally, the releasing Mn2+ exhibits excellent magnetic resonance imaging (MRI) contrast properties. Both in vitro and in vivo studies demonstrate that MnOx-GOx-PM@Ma effectively suppresses tumor growth through synergistic starvation therapy, enhanced CDT, and autophagy inhibition. Collectively, this work presents a strategy to overcome autophagy-mediated therapeutic resistance and optimize synergistic CDT-based antitumor therapy.