Hybrid Techniques for the Diagnosis of Acute Lymphoblastic Leukemia Based on Fusion of CNN Features

Diagnostics, Год журнала: 2023, Номер 13(6), С. 1026 - 1026

Опубликована: Март 8, 2023

Acute lymphoblastic leukemia (ALL) is one of the deadliest forms due to bone marrow producing many white blood cells (WBC). ALL most common types cancer in children and adults. Doctors determine treatment according its stages spread body. rely on analyzing samples under a microscope. Pathologists face challenges, such as similarity between infected normal WBC early stages. Manual diagnosis prone errors, differences opinion, lack experienced pathologists compared number patients. Thus, computer-assisted systems play an essential role assisting detection ALL. In this study, with high efficiency accuracy were developed analyze images C-NMC 2019 ALL-IDB2 datasets. all proposed systems, micrographs improved then fed active contour method extract WBC-only regions for further analysis by three CNN models (DenseNet121, ResNet50, MobileNet). The first strategy two datasets hybrid technique CNN-RF CNN-XGBoost. DenseNet121, MobileNet deep feature maps. produce features redundant non-significant features. So, maps Principal Component Analysis (PCA) select highly representative sent RF XGBoost classifiers classification using serially fused models. DenseNet121-ResNet50, ResNet50-MobileNet, DenseNet121-MobileNet, DenseNet121-ResNet50-MobileNet merged classified XGBoost. classifier reached AUC 99.1%, 98.8%, sensitivity 98.45%, precision 98.7%, specificity 98.85% dataset. With dataset, achieved 100% results AUC, accuracy, sensitivity, precision, specificity.

Язык: Английский

---

Multi-task prediction-based graph contrastive learning for inferring the relationship among lncRNAs, miRNAs and diseases

Briefings in Bioinformatics, Год журнала: 2023, Номер 24(5)

Опубликована: Авг. 1, 2023

Identifying the relationships among long non-coding RNAs (lncRNAs), microRNAs (miRNAs) and diseases is highly valuable for diagnosing, preventing, treating prognosing diseases. The development of effective computational prediction methods can reduce experimental costs. While numerous have been proposed, they often to treat lncRNA-disease associations (LDAs), miRNA-disease (MDAs) lncRNA-miRNA interactions (LMIs) as separate task. Models capable predicting all three simultaneously remain relatively scarce. Our aim perform multi-task predictions, which not only construct a unified framework, but also facilitate mutual complementarity information lncRNAs, miRNAs diseases.In this work, we propose novel unsupervised embedding method called graph contrastive learning (GCLMTP). approach aims predict LDAs, MDAs LMIs by extracting representations To achieve this, first triple-layer lncRNA-miRNA-disease heterogeneous (LMDHG) that integrates complex between these entities based on their similarities correlations. Next, employ an model extract potential topological feature from LMDHG. leverages convolutional network architectures maximize patch corresponding high-level summaries Subsequently, tasks, multiple classifiers are explored LDA, MDA LMI scores. Comprehensive experiments conducted two datasets (from older newer versions database, respectively). results show GCLMTP outperforms other state-of-the-art disease-related lncRNA miRNA tasks. Additionally, case studies further demonstrate ability accurately discover new associations. ensure reproducibility made source code publicly available at https://github.com/sheng-n/GCLMTP.

Язык: Английский

---

Graph Convolutional Network with Neural Collaborative Filtering for Predicting miRNA-Disease Association

Biomedicines, Год журнала: 2025, Номер 13(1), С. 136 - 136

Опубликована: Янв. 8, 2025

Background: Over the past few decades, micro ribonucleic acids (miRNAs) have been shown to play significant roles in various biological processes, including disease incidence. Therefore, much effort has devoted discovering pivotal of miRNAs incidence understand underlying pathogenesis human diseases. However, identifying miRNA-disease associations using experiments is inefficient terms cost and time. Methods: Here, we discuss a novel machine-learning model that effectively predicts disease-related graph convolutional neural network with collaborative filtering (GCNCF). By applying network, could capture important feature vectors present while preserving structure. exploiting filtering, were learned through matrix factorization deep learning, identified. Results: Extensive experimental results based on area under curve (AUC) scores (0.9216 0.9018) demonstrated superiority our over previous models. Conclusions: We anticipate not only serve as an effective tool for predicting but be employed universal computational framework inferring relationships across entities.

Язык: Английский

---

Predicting miRNA-disease association via graph attention learning and multiplex adaptive modality fusion

Computers in Biology and Medicine, Год журнала: 2023, Номер 169, С. 107904 - 107904

Опубликована: Дек. 28, 2023

Язык: Английский

---

MGATAF: multi-channel graph attention network with adaptive fusion for cancer-drug response prediction

BMC Bioinformatics, Год журнала: 2025, Номер 26(1)

Опубликована: Янв. 17, 2025

Drug response prediction is critical in precision medicine to determine the most effective and safe treatments for individual patients. Traditional methods relying on demographic genetic data often fall short accuracy robustness. Recent graph-based models, while promising, frequently neglect role of atomic interactions fail integrate drug fingerprints with SMILES comprehensive molecular graph construction. We introduce multimodal multi-channel attention network adaptive fusion (MGATAF), a framework designed enhance predictions by capturing both local global among nodes. MGATAF improves representation integrating fingerprints, resulting more precise effects. The methodology involves constructing graphs, employing networks capture diverse interactions, using these at multiple abstraction levels. Empirical results demonstrate MGATAF's superior performance compared traditional other techniques. For example, GDSC dataset, achieved 5.12% improvement Pearson correlation coefficient (PCC), reaching 0.9312 an RMSE 0.0225. Similarly, new cell-line tests, outperformed baselines PCC 0.8536 0.0321 0.7364 0.0531 CCLE dataset. significantly advances effectively types complex interactions. This enhances offers robust tool personalized medicine, potentially leading safer Future research can expand this work exploring additional modalities refining mechanisms.

Язык: Английский

---

lncRNA-disease association prediction based on optimizing measures of multi-graph regularized matrix factorization

Computer Methods in Biomechanics & Biomedical Engineering, Год журнала: 2025, Номер unknown, С. 1 - 16

Опубликована: Март 20, 2025

In this paper, we propose a novel lncRNA-disease association prediction algorithm based on optimizing measures of multi-graph regularized matrix factorization (OM-MGRMF). The method first calculates the semantic similarity diseases, functional lncRNAs, and Gaussian both. It then constructs new by using K-nearest-neighbor (KNN) algorithm. Finally, objective function is constructed through utilization ranking regularization constraints. This iteratively optimized an adaptive gradient descent experimental results OM-MGRMF outperform those classical methods in both K-fold cross-validation.

Язык: Английский

---

MUSCLE: multi-view and multi-scale attentional feature fusion for microRNA–disease associations prediction

Briefings in Bioinformatics, Год журнала: 2024, Номер 25(3)

Опубликована: Март 27, 2024

Abstract MicroRNAs (miRNAs) synergize with various biomolecules in human cells resulting diverse functions regulating a wide range of biological processes. Predicting potential disease-associated miRNAs as valuable biomarkers contributes to the treatment diseases. However, few previous methods take holistic perspective and only concentrate on isolated miRNA disease objects, thereby ignoring that are responsible for multiple relationships. In this work, we first constructed multi-view graph based relationships between biomolecules, then utilized attention neural network learn topology features diseases each view. Next, added an mechanism again, developed multi-scale feature fusion module, aiming determine optimal results addition, prior attribute knowledge was simultaneously achieve better prediction solve cold start problem. Finally, learned representations were concatenated fed into multi-layer perceptron end-to-end training predicting miRNA–disease associations. To assess efficacy our model (called MUSCLE), performed 5- 10-fold cross-validation (CV), which got average Area under ROC curves 0.966${\pm }$0.0102 0.973${\pm }$0.0135, respectively, outperforming most current state-of-the-art models. We examined impact crucial parameters performance ablation experiments combination architecture. Furthermore, case studies about colon cancer, lung cancer breast also fully demonstrate good inductive capability MUSCLE. Our data code free available at public GitHub repository: https://github.com/zht-code/MUSCLE.git.

Язык: Английский

---

Establishing a GRU-GCN coordination-based prediction model for miRNA-disease associations

BMC Genomic Data, Год журнала: 2025, Номер 26(1)

Опубликована: Янв. 14, 2025

miRNAs (microRNAs) are endogenous RNAs with lengths of 18 to 24 nucleotides and play critical roles in gene regulation disease progression. Although traditional wet-lab experiments provide direct evidence for miRNA-disease associations, they often time-consuming complicated analyze by current bioinformatics tools. In recent years, machine learning (ML) deep (DL) techniques powerful tools large-scale biological data. Hence, developing a model predict, identify, rank connections diseases can significantly enhance the precision efficiency investigating relationships between diseases. this study, we utilized association data obtained biotechnological develop DL associations. To improve accuracy prediction model, introduced two labeling strategies, weight-based majority-based definitions, classify After preprocessing, was trained novel combining gated recurrent units (GRU) graph convolutional network (GCN) predict level The datasets were from HMDD (the Human miRNA Disease Database) categorized distinct approaches, definitions definitions. We classified associations into three groups, "upregulated", "downregulated" "nonspecific", regression analysis multiclass classification. This GRU-GCN coordinated achieved robust area under curve (AUC) score 0.8 all datasets, demonstrating efficacy predicting potential relationships. By introducing innovative label-preprocessing methods, study addressed diseases, improved ambiguity results different experiments. Based on these refined label developed DL-based refine offers valuable tool complementing experimental methods enhancing our understanding miRNA-related mechanisms.

Язык: Английский

---

DeepWalk-Based Graph Embeddings for miRNA–Disease Association Prediction Using Deep Neural Network

Biomedicines, Год журнала: 2025, Номер 13(3), С. 536 - 536

Опубликована: Фев. 20, 2025

Background: In recent years, micro ribonucleic acids (miRNAs) have been recognized as key regulators in numerous biological processes, particularly the development and progression of diseases. As a result, extensive research has focused on uncovering critical involvement miRNAs disease mechanisms to better comprehend underlying causes human Despite these efforts, relying solely experiments identify miRNA-disease associations is both time-consuming costly, making it an impractical approach for large-scale studies. Methods: this paper, we propose novel DeepWalk-based graph embedding method predicting miRNA–disease association (DWMDA). Using DeepWalk, extracted meaningful low-dimensional vectors from miRNA networks. Then, applied deep neural network using diseases via DeepWalk. Results: An ablation study was conducted assess proposed modules. Furthermore, DWMDA demonstrates exceptional performance two major cancer case studies (breast lung), with results based statistically robust measures, further emphasizing its reliability identifying between Conclusions: We expect that our model will not only facilitate accurate prediction disease-associated but also serve generalizable framework exploring interactions among various entities.

Язык: Английский

---

Identification of UBE2N as a biomarker of Alzheimer’s disease by combining WGCNA with machine learning algorithms

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Фев. 22, 2025

Alzheimer's disease (AD) is the most common cause of dementia, emphasizing critical need for development biomarkers that facilitate accurate and objective assessment progression early detection intervention to delay its onset. In our study, three AD datasets from Gene Expression Omnibus (GEO) database were integrated differential expression analysis, followed by a weighted gene co-expression network analysis (WGCNA), potential screened. Our study identified UBE2N as promising biomarker AD. Functional enrichment revealed associated with synaptic vesicle cycling T cell/B cell receptor signaling pathways. Notably, levels found be significantly reduced in cortex hippocampus TauP301S mice. Furthermore, single-cell data patients demonstrated association function. These findings underscore valuable AD, offering important insights diagnosis targeted therapeutic strategies.

Язык: Английский

---