Negative logarithm of the acid dissociation constant (pK _a) significantly influences the absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of molecules and is a crucial indicator in drug research. Given the rapid and accurate characteristics of computational methods, their role in predicting drug properties is increasingly important. Although many pK _a prediction models currently exist, they often focus on enhancing model precision while neglecting interpretability. In this study, we present GraFpK _a, a pK _a prediction model using graph neural networks (GNNs) and molecular fingerprints. The results show that our acidic and basic models achieved mean absolute errors (MAEs) of 0.621 and 0.402, respectively, on the test set, demonstrating good predictive performance. Notably, to improve interpretability, GraFpK _a also incorporates Integrated Gradients (IGs), providing a clearer visual description of the atoms significantly affecting the pK _a values. The high reliability and interpretability of GraFpK _a ensure accurate pK _a predictions while also facilitating a deeper understanding of the relationship between molecular structure and pK _a values, making it a valuable tool in the field of pK _a prediction.

Negative logarithm of the acid dissociation constant(pKa)significantly influences the absorption,dis-tribution,metabolism,excretion,and toxicity(ADMET)properties of molecules and is a crucial indicator in drug research.Given the rapid and accurate characteristics of computational methods,their role in predicting drug properties is increasingly important.Although many pKa prediction models currently exist,they often focus on enhancing model precision while neglecting interpretability.In this study,we present GraFpKa,a pKa prediction model using graph neural networks(GNNs)and molecular finger-prints.The results show that our acidic and basic models achieved mean absolute errors(MAEs)of 0.621 and 0.402,respectively,on the test set,demonstrating good predictive performance.Notably,to improve interpretability,GraFpKa also incorporates Integrated Gradients(IGs),providing a clearer visual description of the atoms significantly affecting the pKa values.The high reliability and interpretability of GraFpKa ensure accurate pKa predictions while also facilitating a deeper understanding of the relation-ship between molecular structure and pKa values,making it a valuable tool in the field of pKa prediction.

Accurate prediction of drug-target interactions (DTIs) plays a pivotal role in drug discovery, facilitating optimization of lead compounds, drug repurposing and elucidation of drug side effects. However, traditional DTI prediction methods are often limited by incomplete biological data and insufficient representation of protein features. In this study, we proposed KG-CNNDTI, a novel knowledge graph-enhanced framework for DTI prediction, which integrates heterogeneous biological information to improve model generalizability and predictive performance. The proposed model utilized protein embeddings derived from a biomedical knowledge graph via the Node2Vec algorithm, which were further enriched with contextualized sequence representations obtained from ProteinBERT. For compound representation, multiple molecular fingerprint schemes alongside the Uni-Mol pre-trained model were evaluated. The fused representations served as inputs to both classical machine learning models and a convolutional neural network-based predictor. Experimental evaluations across benchmark datasets demonstrated that KG-CNNDTI achieved superior performance compared to state-of-the-art methods, particularly in terms of Precision, Recall, F1-Score and area under the precision-recall curve (AUPR). Ablation analysis highlighted the substantial contribution of knowledge graph-derived features. Moreover, KG-CNNDTI was employed for virtual screening of natural products against Alzheimer's disease, resulting in 40 candidate compounds. 5 were supported by literature evidence, among which 3 were further validated in vitro assays.
Alzheimer Disease/drug therapy* ; Biological Products/therapeutic use* ; Humans ; Neural Networks, Computer ; Machine Learning ; Drug Discovery/methods* ; Algorithms ; Drug Evaluation, Preclinical/methods*

Objective:To summarize the clinical characteristics of focal cerebral arteriopathy (FCA) in children, and to analyze its influencing factor of prognosis.Methods:A retrospective cohort study was conducted. Clinical data from 40 children with FCA who were hospitalized at the Department of Neurology, Beijing Children′s Hospital, Capital Medical University, from September 2015 to August 2024 were collected. A centralized follow-up was conducted in October 2024 via outpatient clinics or the internet. The pediatric stroke outcome measure (PSOM) was used to evaluate their outcomes. Based on the PSOM, the children were further divided into a group with normal neurological function and another group with abnormal neurological function. Differences between groups were analyzed using the Mann-Whitney U test and Fisher exact test. Univariate Logistic regression analysis was performed to identify the influencing factors for neurological outcomes in children with FCA. Results:A total of 40 children were included, with 20 males and 20 females, and the onset age of 9.2 (6.8, 12.5) years. Among them, 12 cases (30%) had a history of varicella within 1 year before onset. There were 23 cases (58%) presenting with transient ischemic attack (TIA) or recurrent fluctuating symptoms of onset, while 3 cases (8%) developed progressive stroke within the first month of onset. The M1 segment of the middle cerebral artery was the most commonly affected vascular site, with a total of 16 cases (40%). Arterial occlusion occurred in 8 cases (20%). Lumbar puncture was completed in 36 children, and white blood cell counts in cerebrospinal fluid was increased in 6 cases. All 23 patients who completed magnetic resonance vessel wall imaging (VWI) showed circular enhancement of the arterial wall. A total of 28 patients (70%) received antiplatelet or anticoagulation therapy, and 16 patients (40%) received hormone therapy. At admission, the pediatric National Institute of Health Stroke Scale (PedNIHSS) score was 6.0 (2.0, 8.8) points, which decreased to 0.5 (0, 3.0) points at discharge. The follow-up duration was 1.6 (0.8, 4.9) years, with 1 case lost to follow-up. There was 1 case presenting with recurrence course manifesting as TIA. Among the 39 cases who completed the follow-up, 23 cases (59%) were assessed as neurologically normal by PSOM, while 16 cases (41%) were assessed as neurologically abnormal. Among the 29 cases who completed the imaging review, magnetic resonance angiography (MRA) review in 23 cases indicated stability or improvement in the original arterial stenosis, with 6 cases experiencing transient worsening of arterial stenosis early in the disease course (within 2 months), which later improved. Arterial stenosis progression occurred in 6 cases at the final review of 29 cases who completed the imaging review, with 1 case developing progressive cerebral arteriopathy. The proportion of patients with headache, altered consciousness, and aphasia in the abnormal neurological function group, as well as the PedNISS scores at admission and discharge, were all higher than those in the normal neurological function group (all P<0.05). Univariate Logistic regression analysis revealed that only a PedNISS score>6 points at onset was an influencing factor for abnormal neurological function ( OR=20.58, 95% CI 3.93-107.70, P<0.001). Conclusions:Childhood FCA often presents with fluctuating onset, and the proximal segment of the middle cerebral artery is frequently affected. Progression of arterial stenosis is common within 2 months of the disease course, but clinical progression and new ischemic lesions are uncommon. Most patients have a favorable long-term prognosis. PedNIHSS score>6 points at admission is related to abnormal neurological function outcomes.

Objective:To summarize the long-term efficacy of nicotinamide in treating pediatric early-onset progressive encephalopathy with brain edema and (or) leukoencephalopathy-2 (PEBEL2) caused by NAXD gene variation .Methods:This was a case report conducted from February 2019 to January 2025. The long-term efficacy of nicotinamide was observed by following up a child with PEBEL2 who received the treatment in the Department of Neurology, Beijing Children′s Hospital Affiliated to Capital Medical University. The clinical data included changes in skin lesions, neurological symptoms. The modified Rankin scale (mRS) was used to evaluate the recovery of neurological function.Results:A boy was diagnosed with PEBEL2 caused by NAXD gene variation via genetic testing at Beijing Children′s Hospital Affiliated to Capital Medical University in February 2019, when he was 4 years and 6 months of age. Immediately after diagnosis, nicotinamide treatment was initiated at an initial dose of 100 mg/d, which was increased by 100 mg per week and gradually increased to 500 mg/d; meanwhile, other therapeutic drugs were gradually discontinued. After 1 year and 8 months of treatment, the child′s skin lesions had completely resolved; at the 2-year follow-up, dystonia in both upper limbs and swallowing dysfunction was alleviated significantly; by 2.5-year follow-up, his cognitive function also showed improvement. When the child was treated with 500 mg/d for 3 years, a rash appeared around the mouth. After the dose was reduced to 250 mg/d, the rash resolved, and the dose of 250 mg/d was maintained until the last follow-up. At the last follow-up in January 2025, the child was 10 years and 5 months of age. His mRS score decreased from 5 (before treatment) to 4. During the 6-year of continuous nicotinamide treatment, the child′s condition remained stable without progression. Drug-related skin rashes occurred, but no severe drug-related adverse reactions were observed.Conclusions:PEBEL2 is a treatable mitochondrial disease. Nicotinamide treatment can effectively improve skin lesions and neurological symptoms in PEBEL2 patients, and the long-term administration demonstrates a favorable safety profile.

Objective:To analyze the clinical characteristics, RAN-binding protein 2 ( RANBP2) gene variations, and prognosis in Chinese acute necrotizing encephalopathy (ANE) patients. Methods:A retrospective analysis of ANE cases registered in the Peking Union Medical College Hospital Encephalitis Registry System from 2022 to 2024, involving patients from Peking Union Medical College Hospital and other hospitals, was conducted. A descriptive study was performed on the clinical characteristics, treatments and prognosis, cerebrospinal fluid examination results, and imaging findings of these patients based on adjusted ANE diagnostic criteria. Whole-exome sequencing technology was used to detect gene mutations in these patients.Results:A total of 18 ANE cases were included, ranged in age from 2 to 72 [20(5, 43)] years. The male-to-female ratio was 4∶5. All patients were found with precipitating infections including COVID-19, influenza A virus and Mycoplasma pneumoniae infections. All patients presented with fever, with varying degrees of consciousness disturbance observed in 16 cases, and seizures in 10 cases. All patients underwent lumbar puncture, with normal or mildly elevated white cell counts [3(2, 13)×10 6/L] and mildly to moderately elevated protein levels [1.90(0.92, 4.65) g/L]. A total of 6 patients were found with extremely elevated interleukin-6 level [950(164, 2 000) pg/ml] in cerebrospinal fluid. Bilateral symmetric thalamic lesions were typical imaging features of ANE, while involvement of other areas such as cortical and subcortical white matter, brainstem, and cerebellum was also observed. A total of 14 patients performed genetic tests while 4 patients were identified with RANBP2 gene mutations (c.1754C>T in 3 cases, c.1966A>G in 1 case). All patients received immunotherapy, and 7 patients died at discharge while other patients presented with neurological sequelae of varying degrees. Conclusions:ANE is a rare and severe parainfectious encephalopathy that can occur in both children and adults. Clinically, it is characterized by rapidly progressing encephalopathy following systematic infection, with bilateral symmetric thalamic lesions. The detection of RANBP2 gene mutations could help make the diagnosis.

Background and purpose:Melanoma is a highly invasive malignant tumor originating from melanocytes,which poses a great threat to human life and health around the world,and its morbidity and mortality have been rising continuously in recent years.Telomerase and autophagy play crucial roles in cell proliferation,survival and stress response.Telomerase maintains the replication ability of cells by prolonging telomeres at the ends of chromosomes,and autophagy,as a self-degradation mechanism of cells,can not only help cells remove damaged components to promote survival,but also induce cell death under certain conditions.In the tumor environment,they are often abnormally activated or out of balance,and participate in the occurrence and development of many cancers,including melanoma.This study investigated the roles of telomerase and autophagy in melanoma progression and evaluated the potential synergistic therapeutic effects of combined application of telomerase inhibitor BIBR1532 and autophagy inhibitor chloroquine(CQ)in melanoma treatment.Methods:Malignant melanoma cells A375 were treated with telomerase inhibitor BIBR1532.The cell viability was assessed using the cell counting kit-8(CCK-8)assay,and the cell apoptosis was detected using the Annexin Ⅴ/propidium iodide(PI)double staining method.Additionally,the expressions of autophagy-related proteins LC3-Ⅱand p62 were detected by Western blot,and the changes in autophagy flux were observed using dual-tagged adenovirus transfection technology.Based on these studies,BIBR1532 and the autophagy inhibitor CQ were further applied in combination to analyze cell proliferation,apoptotic rate,changes in mitochondrial membrane potential,and cell cycle distribution,and the cloning formation experiment was used to verify the cell's proliferative capacity,thereby comprehensively evaluating the efficacy of this combined treatment strategy.Results:Telomerase inhibitor BIBR1532 at a concentration of 50 μmol/L significantly inhibited the growth of malignant melanoma cells A375 and induced apoptosis.At the same concentration,BIBR1532 upregulated the expression of the autophagy-related protein LC3-Ⅱ in A375 cells,while downregulating the expression of p62 protein.By transducing A375 cells with a dual-tagged adenovirus,it was observed that autophagy flux was significantly enhanced after treatment with BIBR1532.Furthermore,the combined application of BIBR1532(50 μmol/L)and the autophagy inhibitor CQ(20 μmol/L)significantly promoted the death of A375 cells,induced apoptosis and destruction of mitochondrial membrane potential,caused cell cycle arrest at the G2/M phase,and significantly inhibited the cell's clonogenic ability.Conclusion:Telomerase inhibitor BIBR1532 not only inhibits the proliferation of malignant melanoma cells but also activates the autophagy process in these cells,and inhibition of the autophagy response by autophagy inhibitor CQ can enhance the sensitivity of malignant melanoma cells to telomerase inhibitor BIBR1532.

In recent years, generative artificial intelligence (AI) technologies have achieved remarkable progress in the early screening, risk prediction, disease progression assessment, and clinical trial design of glaucoma.Using advanced algorithms, such as generative adversarial networks, variational autoencoders, and diffusion models, researchers have synthesized high-quality structural images of the optic disc, macular region, and retinal nerve fiber layer, which effectively alleviates the limitations of scarce clinical imaging data and label imbalance.These methods have substantially improved the accuracy and generalization of deep learning models in visual field defect prediction, structure-function mapping, and longitudinal disease progression simulation.Meanwhile, multimodal generative approaches that integrate imaging data, visual field tests, and clinical features have facilitated individualized prediction of glaucoma progression.In addition, large language models have shown preliminary potential in ophthalmic image interpretation, clinical text information extraction, and decision support, providing new insights into intelligent ophthalmic diagnosis and treatment.However, the clinical implementation of generative AI in glaucoma faces challenges.The pathological authenticity and cross-device consistency of generated images require further validation, which may affect the reliability of early glaucoma detection.The heterogeneous characteristics of different glaucoma subtypes, such as open-angle and angle-closure glaucoma, also limit the generalization of synthetic data.Moreover, issues related to model interpretability (" black-box" nature), artifact generation, data privacy, and ethical governance remain key barriers to clinical translation.In the future, it is expected that establishing large-scale training frameworks that incorporate multicenter, multimodal, and multiethnic datasets will enhance model robustness and clinical applicability.Furthermore, generative AI may contribute to remote ophthalmic care and personalized precision therapy by enhancing low-quality image, reconstructing missing data, and simulating dynamic disease courses.This article reviews the current applications, core technologies, and challenges of generative AI in glaucoma diagnosis and management, and discusses its future directions and translational potential in clinical ophthalmology.

In recent years, generative artificial intelligence (AI) technologies have achieved remarkable progress in the early screening, risk prediction, disease progression assessment, and clinical trial design of glaucoma.Using advanced algorithms, such as generative adversarial networks, variational autoencoders, and diffusion models, researchers have synthesized high-quality structural images of the optic disc, macular region, and retinal nerve fiber layer, which effectively alleviates the limitations of scarce clinical imaging data and label imbalance.These methods have substantially improved the accuracy and generalization of deep learning models in visual field defect prediction, structure-function mapping, and longitudinal disease progression simulation.Meanwhile, multimodal generative approaches that integrate imaging data, visual field tests, and clinical features have facilitated individualized prediction of glaucoma progression.In addition, large language models have shown preliminary potential in ophthalmic image interpretation, clinical text information extraction, and decision support, providing new insights into intelligent ophthalmic diagnosis and treatment.However, the clinical implementation of generative AI in glaucoma faces challenges.The pathological authenticity and cross-device consistency of generated images require further validation, which may affect the reliability of early glaucoma detection.The heterogeneous characteristics of different glaucoma subtypes, such as open-angle and angle-closure glaucoma, also limit the generalization of synthetic data.Moreover, issues related to model interpretability (" black-box" nature), artifact generation, data privacy, and ethical governance remain key barriers to clinical translation.In the future, it is expected that establishing large-scale training frameworks that incorporate multicenter, multimodal, and multiethnic datasets will enhance model robustness and clinical applicability.Furthermore, generative AI may contribute to remote ophthalmic care and personalized precision therapy by enhancing low-quality image, reconstructing missing data, and simulating dynamic disease courses.This article reviews the current applications, core technologies, and challenges of generative AI in glaucoma diagnosis and management, and discusses its future directions and translational potential in clinical ophthalmology.

Background and purpose:Melanoma is a highly invasive malignant tumor originating from melanocytes,which poses a great threat to human life and health around the world,and its morbidity and mortality have been rising continuously in recent years.Telomerase and autophagy play crucial roles in cell proliferation,survival and stress response.Telomerase maintains the replication ability of cells by prolonging telomeres at the ends of chromosomes,and autophagy,as a self-degradation mechanism of cells,can not only help cells remove damaged components to promote survival,but also induce cell death under certain conditions.In the tumor environment,they are often abnormally activated or out of balance,and participate in the occurrence and development of many cancers,including melanoma.This study investigated the roles of telomerase and autophagy in melanoma progression and evaluated the potential synergistic therapeutic effects of combined application of telomerase inhibitor BIBR1532 and autophagy inhibitor chloroquine(CQ)in melanoma treatment.Methods:Malignant melanoma cells A375 were treated with telomerase inhibitor BIBR1532.The cell viability was assessed using the cell counting kit-8(CCK-8)assay,and the cell apoptosis was detected using the Annexin Ⅴ/propidium iodide(PI)double staining method.Additionally,the expressions of autophagy-related proteins LC3-Ⅱand p62 were detected by Western blot,and the changes in autophagy flux were observed using dual-tagged adenovirus transfection technology.Based on these studies,BIBR1532 and the autophagy inhibitor CQ were further applied in combination to analyze cell proliferation,apoptotic rate,changes in mitochondrial membrane potential,and cell cycle distribution,and the cloning formation experiment was used to verify the cell's proliferative capacity,thereby comprehensively evaluating the efficacy of this combined treatment strategy.Results:Telomerase inhibitor BIBR1532 at a concentration of 50 μmol/L significantly inhibited the growth of malignant melanoma cells A375 and induced apoptosis.At the same concentration,BIBR1532 upregulated the expression of the autophagy-related protein LC3-Ⅱ in A375 cells,while downregulating the expression of p62 protein.By transducing A375 cells with a dual-tagged adenovirus,it was observed that autophagy flux was significantly enhanced after treatment with BIBR1532.Furthermore,the combined application of BIBR1532(50 μmol/L)and the autophagy inhibitor CQ(20 μmol/L)significantly promoted the death of A375 cells,induced apoptosis and destruction of mitochondrial membrane potential,caused cell cycle arrest at the G2/M phase,and significantly inhibited the cell's clonogenic ability.Conclusion:Telomerase inhibitor BIBR1532 not only inhibits the proliferation of malignant melanoma cells but also activates the autophagy process in these cells,and inhibition of the autophagy response by autophagy inhibitor CQ can enhance the sensitivity of malignant melanoma cells to telomerase inhibitor BIBR1532.