Find the papers that actually matter

Ubiquitin-specific protease 7 (USP7) is a key member of the deubiquitinating enzyme family. It is abnormally overexpressed in various malignancies, including breast cancer, chronic lymphocytic leukemia, and prostate cancer. By regulating pathways such as the p53-MDM2 signaling axis, USP7 promotes tumorigenesis and progression, making it a highly promising therapeutic target for anticancer treatment. Although multiple USP7 inhibitors have been reported, existing screening and evaluation assays exhibit limitations: the ubiquitin-phospholipase A2 (Ub-PLA2) assay frequently produces false-positive results, while the ubiquitin-rhodamine (Ub-Rho) assay is susceptible to interference from compound autofluorescence. To address this challenge, we developed a fluorescence polarization (FP) assay. This employs a rationally designed strategy that exhibits excellent characteristics, making it a simple-to-operate and cost-effective method, suitable for the evaluation of compound bioactivity against USP7. To further validate the practicality and reliability of this FP assay, we conducted a structure-based drug design campaign involving two rounds of systematic structural optimization, yielding 51 novel derivatives featuring pyrazolo[4,3-d]pyrimidine and piperidol scaffolds. Following FP evaluation and Ub-Rho enzyme activity validation, we performed a comprehensive structure-activity relationship (SAR) analysis. Ultimately, in vitro cellular assays identified three compounds (LC-U7-44, LC-U7-48, and LC-U7-50) that exhibit potent USP7 inhibitory activity alongside favorable cellular anti-proliferative effects. Overall, the established FP assay in this study closes a methodological gap in the evaluation of USP7 inhibitors, and the detailed SAR analysis provides a foundation for the further development of potent USP7 inhibitors.

Triple-negative breast cancer (TNBC) is an aggressive subtype lacking the ER, PR, and HER2 receptors, with limited treatment options and poor prognosis. Ferroptosis, an iron-dependent form of regulated cell death driven by lipid peroxidation, has emerged as a promising therapeutic strategy for TNBC. Glutathione peroxidase 4 (GPX4) is a key ferroptosis suppressor, and its inhibition sensitizes TNBC cells to oxidative damage. To discover and characterize DA-5, a novel 3,5-disubstituted azaindole derivative inspired by compounds from the traditional Chinese medicine Shuganning injection (SGNI), as a potent GPX4 inhibitor to induce ferroptosis in TNBC cells. Molecular docking and structural optimization were used to design DA-5. Its binding affinity (Kd) and enzymatic inhibitory activity (IC50) against GPX4 were evaluated. In vitro assays assessed DA-5's ability to induce ferroptosis in TNBC cells through lipid peroxidation while sparing normal mammary epithelial cells. In vivo studies evaluated the efficacy and safety of DA-5 in TNBC xenograft models via oral administration. Pharmacokinetic profiles were also analyzed. DA-5 demonstrated high-affinity binding to GPX4 (Kd = 10.04 μM) and effectively suppressed its enzymatic activity (IC50 = 10.90 μM). In TNBC cells, DA-5 promoted lipid peroxidation and induced ferroptosis, which was rescued by ferroptosis inhibitors and by iron chelators. Oral administration of DA-5 significantly inhibited TNBC xenograft growth in vivo without systemic toxicity, supported by favorable pharmacokinetic and safety profiles. These findings identify DA-5 as a novel azaindole-based GPX4 inhibitor capable of inducing ferroptosis through GPX4-targeted lipid peroxidation. This breakthrough offers a promising targeted therapy for TNBC.

Accurate subtyping of lung cancer is essential for improving patient prognosis and enabling personalized treatment. However, current clinical techniques are often time-consuming and heavily dependent on the operator's subjective judgment and experience, which limits the accuracy and timeliness of intraoperative subtype diagnosis and margin assessment. In this study, we developed an intelligent diagnostic model by integrating Fourier transform infrared (FTIR) spectroscopy with a Random Forest (RF) classifier. A total of 210 tumor and adjacent tissue samples from 105 patients, including adenocarcinoma, squamous cell carcinoma, and benign lung tumors were analyzed. The constructed RF model achieved an accuracy of 97.95% with an Area Under the Curve (AUC) of 0.99 in binary classification (lung cancer vs. adjacent tissues), and an accuracy of 94.91% in multiclass classification of lung cancer subtypes, significantly outperforming conventional algorithms such as Support Vector Machine, Naive Bayes, and Logistic Regression. In addition, spectral analysis methods, including peak area comparison, peak fitting, and second derivative analysis, revealed distinct differences in nucleic acids, proteins, and lipids, highlighting the characteristic bands responsible for subtype discrimination and providing spectroscopic insights into the pathological features of different lung cancer subtypes. Collectively, our findings demonstrate that the diagnostic model is a powerful approach for distinguishing lung cancer tissues from normal tissues and for subtype classification, offering a promising tool for lung cancer diagnosis.

Artificial intelligence-based radiomics offers a potential adjunct to the current clinical management of paediatric brain tumours by enabling prediction of key diagnostic features. However, despite promising research, integration into clinical practice remains limited. This may in part reflect limitations in the underlying evidence base. Previous research has identified inconsistent methodological quality in primary studies and systematic reviews, as well as inadequate risk of bias assessment of primary studies within reviews. Given these concerns, rigorous appraisal of the available evidence base is essential to determine whether current findings are reliable and suitable for clinical translation. This study aimed to characterise the methodological appraisal approaches used in systematic reviews of artificial intelligence-based radiomics research in paediatric brain tumour imaging, with particular attention to how review authors select and apply tools for risk of bias, methodological quality, and reporting completeness. We performed a meta-research study of systematic reviews. PubMed, Embase, Web of Science, Scopus and Medline were searched in March 2024 without date limits. Study selection followed PRISMA 2020 guidelines. Reviews of artificial intelligence-based radiomics applications in predominately paediatric populations with primary brain tumours were included. We evaluated the methodological appraisal approaches used, including tools assessing risk of bias, methodological quality and reporting completeness. Seven systematic reviews met inclusion criteria. Among these, three reviews (3/7) did not employ any formal methodological appraisal tool, and only one provided a rationale for this omission. Four reviews (4/7) applied at least one formal appraisal tool using five distinct instruments in total. Two (2/5) of these instruments were designed specifically for artificial intelligence-based research. The Quality Assessment of Diagnostic Accuracy Studies 2 was the most used tool, applied across two reviews. Methodological appraisal is applied inconsistently in systematic reviews of artificial intelligence-based radiomics research in paediatric neuro-oncology. This may reduce confidence in the current evidence base and hinder clinical translation. While some reviews used formal appraisal tools, these were usually conventional tools that may not fully capture methodological considerations specific to artificial intelligence-based research. Greater use of consistent, transparent, and artificial intelligence-aware appraisal approaches, alongside improved reporting in primary studies, is needed to support more reliable evidence synthesis and translation of radiomics into clinical practice.

Accurate classification of Renal Cell Carcinoma (RCC) subtypes is essential for personalized therapy, as prognostic outcomes and treatment responses differ markedly among subtypes. Most artificial intelligence applications in RCC focus on histology, while immunohistochemistry applications mainly quantify biomarkers rather than perform subtype classification. In clinical diagnostics, immunohistochemistry remains a valuable complement to histology, though its implementation is limited by cost, labor intensity, and resource constraints. This study aims to develop an uncertainty-aware artificial intelligence framework that integrates histological and immunohistochemistry data to improve RCC subtype classification and optimize laboratory workflows. We designed a hierarchical, pathologist-guided artificial intelligence framework that integrates uncertainty estimation into RCC subtype classification. High-confidence histological predictions produced by deep learning models are accepted directly, while low-confidence cases automatically trigger targeted immunohistochemistry analysis automatically analyzed using dedicated machine learning algorithms. To address staining variability across institutions, a CycleGAN-based stain transfer module was employed to harmonize color domains and enhance generalizability. The framework was evaluated on multi-center datasets encompassing different staining protocols. The proposed integrated framework demonstrated significant diagnostic performance improvements. Patient-level accuracy reached 97.5% on internal cross-validation and 95% on external cohorts when selective immunohistochemistry refinement was applied. The uncertainty-driven immunohistochemistry module reduced redundant biomarker testing to approximately one-fourth of all cases while maintaining or improving classification confidence. The stain transfer module also effectively mitigated inter-laboratory color discrepancies, supporting consistent model performance across different centers. Our framework combines histology and immunohistochemistry to deliver accurate and efficient subtype classification. By selectively activating immunohistochemistry analysis for low-confidence predictions on histology, the approach optimizes biomarker use and laboratory resources while maintaining high diagnostic reliability. The study highlights the potential of combining deep learning-based histology with targeted immunohistochemistry biomarker assessment to advance precision and reproducibility in cancer diagnostics workflows.

Large language models (LLMs) have shown promise in medical imaging, but their utility in cytology remains underexplored. This study evaluates GPT-5 and Gemini 2.5 Pro for cervical Pap test interpretation. Digital cervical Pap test images of 100 cases were obtained from the Hologic Education Site, with Hologic diagnoses considered the gold standard. Representative images were uploaded into GPT-5 and Gemini 2.5 Pro and prompted to provide a diagnosis based on the Third Edition of the Bethesda System for Reporting Cervical Cytopathology. Cases with infectious organisms were assessed using additional images. Concordance was evaluated at exact diagnosis and clinical management groupings, wherein diagnoses with similar management implications were grouped. Sensitivity and specificity for abnormal cytology were also calculated. Concordance of both LLMs for exact diagnosis were comparable (GPT-5: 47%, Gemini: 48%) and increased to 66% for clinical management grouping. GPT-5 performed best for low-grade squamous intraepithelial lesions (75%), whereas Gemini 2.5 Pro showed the highest concordance in the high-grade squamous intraepithelial lesion (HSIL) category (82%), although this was largely attributable to its strong tendency to overcall cases as HSIL. Sensitivity for detecting abnormal cytology was 74% for GPT-5 and 84% for Gemini, with specificity of 74% and 71%, respectively. GPT-5 better identified glandular lesions, while Gemini detected organisms more accurately (71% vs. 20%). Current LLMs demonstrate moderate ability to identify cytologic abnormalities but are not yet reliable for independent cervical Pap test interpretation. Fine-tuning, prompt optimization, and cytology-specific training could enhance their utility as adjunctive tools in cytology workflows.

Accurate and up-to-date anatomical information is critical for effective treatment planning in breast cancer adaptive radiotherapy. Cone-beam computed tomography facilitates real-time plan optimization but lacks sufficient electron density accuracy for direct clinical application. To address this limitation, we propose a novel unsupervised deep learning framework that integrates the Mamba architecture with an artifact disentanglement network to form the Artifact Disentanglement Network-Mamba model. This study proposes an unsupervised deep learning framework, ADN-Mamba, integrating an Artifact Disentanglement Network (ADN) with the structured state-space model Mamba for high-precision sCT synthesis from breath-hold CBCT (BH-CBCT). The model uses three encoders (CBCT content, CT content, artifact) and two generators to disentangle anatomical features from artifacts in CBCT. Mamba enhances the ability of the model to capture long-range dependencies, improving representation of complex anatomical structures. The Artifact Disentanglement Network-Mamba model achieved a mean absolute error of 54.97 HU within the body. The mean absolute percent errors of synthetic and real CT images in the soft tissue (-150 HU to 150 HU) and bone (200 HU to 1500 HU) regions were 46.26% and 30.98%, respectively. The gamma pass rate of the calculated dose on sCT compared with that on pCT is 97.74% under the 2%/2 mm criterion. The proposed model outperforms six other state-of-the-art methods in terms of image quality, dose accuracy, and radiomic feature consistency. By overcoming challenges such as registration errors and the absence of paired cone-beam computed tomography-computed tomography datasets, the proposed framework demonstrated superior performance in terms of anatomical fidelity and dose calculation accuracy. ADN-Mamba enables precise BH-CBCT-to-CT synthesis via unsupervised artifact disentanglement and Mamba's long-range modeling, demonstrating superior performance in image quality, dose calculation accuracy, and radiomic consistency. This framework provides a reliable tool for online dose calculation and target delineation in breast ART. Future work will focus on extending the model to 3D data and multicenter validation.

White blood cells (WBCs) and their subpopulations play critical roles in detecting blood cancers due to their distinct biological and biochemical characteristics. Infrared (IR) spectroscopy offers a rapid, label-free, and non-destructive approach to probe molecular composition, making it a promising tool for biomedical diagnostics. The objective of this proof-of-principle study is to investigate the possibility of IR spectroscopy combined with chemometrics to differentiate leukemia from lymphoma, and to assess the capability of whole WBCs and their subpopulations in distinguishing the two diseases. We based our study on 21 pediatric patients including 11 leukemia and 10 lymphoma cases, with in total 86,016 IR spectra measured from whole WBCs and the subpopulations. Data pipeline was established, including steps of spectral preprocessing, classification, and data fusion. Particularly, data fusion was implemented via low-, middle-, and high-level strategies, with the aim of combining spectra from different cell types and investigating their capability of differentiating the two blood cancers. The classification, both with and without data fusion, was benchmarked via the patient-wise cross-validation. A balanced accuracy of 80.0% was achieved based on IR spectra of whole WBCs. Further improvement was observed when combining whole WBCs and its subpopulations, with the best performance of 90.0% from combining whole WBCs and granulocytes with high-level data fusion strategy. The performance was observed consistent for both linear and nonlinear classifications based on linear discriminant analysis (LDA) and support vector machine (SVM), respectively. The results indicate the promising potential of IR spectroscopy of blood samples to distinguish leukemia and lymphoma with the help of chemometric approaches. Further, WBC subpopulations, particularly granulocytes, were proven to contain complementary information to whole WBCs for differentiating leukemia from lymphoma. This provides critical insights for biomedical practice in blood cancer diagnostics.

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.

Aristolochic acids (AAs) are established human carcinogens strongly associated with upper tract urothelial carcinoma (UTUC). However, the multi-target oncogenic network beyond their genotoxic mechanism remains incompletely elucidated. This study employed an integrated computational approach combining network toxicology, machine learning, molecular docking, and molecular dynamics (MD) simulations to systematically explore the potential molecular mechanisms of AA-induced UTUC. We identified 97 shared potential targets of AAs and UTUC. Enrichment analyses revealed their significant involvement in lipid metabolism, xenobiotic detoxification, and cancer-related pathways such as PI3K-Akt signaling. Topological analysis of the protein-protein interaction network and a nested cross-validation machine learning model highlighted five core genes: CASP3, EGFR, PARP1, PTGS2, and HSP90AA1. Molecular docking predicted high binding affinities of AA with these core targets, particularly for PTGS2 (-9.3 kcal/mol) and EGFR (-8.2 kcal/mol). Subsequent 100-ns MD simulations and Molecular Mechanics/Poisson-Boltzmann Surface Area (MM/PBSA) calculations confirmed the structural stability and spontaneous binding (ΔG_bind = -55.68 kcal/mol) of the AA-EGFR complex. Our multi-omics analysis suggests that AAs may promote UTUC not only via canonical DNA adduct formation but also potentially through direct interactions with key signaling proteins, implicating a synergistic mechanism involving both genotoxic and non-genotoxic pathways. These findings provide a theoretical foundation for novel preventive and therapeutic strategies against AA-associated UTUC.