Scopus İndeksli Yayınlar Koleksiyonu
Permanent URI for this collectionhttps://hdl.handle.net/20.500.12573/395
Browse
Search Results
Article Citation - Scopus: 4RCE-IFE: Recursive Cluster Elimination with Intra-Cluster Feature Elimination(PeerJ Inc., 2025-02-07) Kuzudisli, Cihan; Bakir-Gungor, Burcu; Qaqish, Bahjat; Yousef, MalikConference Object Impact of Gene Duplicate Handling Strategies on Classification Performance and Feature Selection in Gene Expression Data(Institute of Electrical and Electronics Engineers Inc., 2025-09-17) Kuzudisli, Cihan; Qaqish, Bahjat; Gungor, Burcu Bakir; Yousef, MalikConference Object Enhancing Complex Disease Group Scoring with Mirgedinet: A Multi-Algorithm Machine Learning Framework Based on the GSM Approach(IEEE, 2025-06-25) Qumsiyeh, Emma; Bakir-Gungor, Burcu; Yousef, MalikIntegrating biological prior knowledge for disease gene associations has shown significant promise in discovering new biomarkers with potential translational applications. This work investigates the application of a multi-algorithm machine learning framework based on the Grouping-Scoring-Modeling (G-S-M) approach for improving the prediction of complex diseases. The study identifies the primary gene and miRNA interactions in various complex diseases with the help of miRGediNET, which is a machine-learning based tool that integrates data from three biological databases. Traditional methods have only focused on independence between features; the G-S-M method focuses on aggregating genes based on biological interactions, pinpointing the scoring of gene groups for a disease, and modeling its predictive capability using advanced machine learning algorithms. In this research paper, seven algorithms, including Support Vector Machine, Decision Tree, and CatBoost, were applied to eight datasets extracted from the GEO database. This framework proved very robust in ranking gene clusters, thus predicting critical biomarkers while doing 100-fold randomized cross-validation within the evaluation. The results indicate this approach's high potential for refining disease and supporting research for choosing the best algorithm that can provide biological insights and computational advances.Correction Correction: Engineering Novel Features for Diabetes Complication Prediction Using Synthetic Electronic Health Records(Frontiers Media S.A., 2025-08-29) Voskergian, Daniel; Bakir-Gungor, Burcu; Yousef, MalikConference Object TextNetTopics+: Enhancing Text Classification Through Classifier Diversity and Model Ensembling(Springer International Publishing AG, 2025) Voskergian, Daniel; Bakir-Gungor, Burcu; Yousef, MalikTextNetTopics is an innovative text classification framework that integrates topic modeling with feature selection to improve model accuracy and interpretability. Unlike traditional methods that rely on individual words, TextNetTopics selects cohesive topics extracted via Latent Dirichlet Allocation as features for document representation, effectively reducing dimensionality while preserving the semantic structure of the text. This study evaluates the performance of TextNetTopics utilizing multiple machine learning algorithms in the M (Modeling) component, including Random Forest, Support Vector Machine, Gradient Boosting, eXtreme Gradient Boosting, and Logistic Regression. To further enhance classification performance, we introduce TextNetTopics+, an ensemblebased extension that leverages both hard voting and soft voting mechanisms to combine the strengths of multiple classifiers. Comprehensive experiments on the LitCovid and WOS datasets demonstrate that ensemble learning in TextNetTopics + significantly outperforms individual classifiers in TextNetTopics, confirming its effectiveness in improving model robustness and generalization.Article Citation - WoS: 2Citation - Scopus: 4RCE-IFE: Recursive Cluster Elimination With Intra-Cluster Feature Elimination(PeerJ Inc, 2025-02-07) Kuzudisli, Cihan; Bakir-Gungor, Burcu; Qaqish, Bahjat; Yousef, MalikThe computational and interpretational difficulties caused by the ever-increasing dimensionality of biological data generated by new technologies pose a significant challenge. Feature selection (FS) methods aim to reduce the dimension, and feature grouping has emerged as a foundation for FS techniques that seek to detect strong correlations among features and identify irrelevant features. In this work, we propose the Recursive Cluster Elimination with Intra-Cluster Feature Elimination (RCE-IFE) method that utilizes feature grouping and iterates grouping and elimination steps in a supervised context. We assess dimensionality reduction and discriminatory capabilities of RCE-IFE on various high-dimensional datasets from different biological domains. For a set of gene expression, MicroRNA (miRNA) expression, and methylation datasets, the performance of RCE-IFE is comparatively evaluated with RCE-IFE-SVM (the SVM-adapted version of RCE-IFE) and SVM-RCE. On average, RCE-IFE attains an area under the curve (AUC) of 0.85 among tested expression datasets with the fewest features and the shortest running time, while RCE-IFE-SVM (the SVM-adapted version of RCE-IFE) and SVM-RCE achieve similar AUCs of 0.84 and 0.83, respectively. RCE-IFE and SVM-RCE yield AUCs of 0.79 and 0.68, respectively when averaged over seven different metagenomics datasets, with RCE-IFE significantly reducing feature subsets. Furthermore, RCE-IFE surpasses several state-of-the-art FS methods, such as Minimum Redundancy Maximum Relevance (MRMR), Fast Correlation-Based Filter (FCBF), Information Gain (IG), Conditional Mutual Information Maximization (CMIM), SelectKBest (SKB), and eXtreme Gradient Boosting (XGBoost), obtaining an average AUC of 0.76 on five gene expression datasets. Compared with a similar tool, Multi-stage, RCE-IFE gives a similar average accuracy rate of 89.27% using fewer features on four cancer-related datasets. The comparability of RCE-IFE is also verified with other biological domain knowledge-based Grouping-Scoring-Modeling (G-S-M) tools, including mirGediNET, 3Mint, and miRcorrNet. Additionally, the biological relevance of the selected features by RCE-IFE is evaluated. The proposed method also exhibits high consistency in terms of the selected features across multiple runs. Our experimental findings imply that RCE-IFE provides robust classifier performance and significantly reduces feature size while maintaining feature relevance and consistency.Article Citation - Scopus: 3Prediction of Colorectal Cancer Based on Taxonomic Levels of Microorganisms and Discovery of Taxonomic Biomarkers Using the Grouping-Scoring (G-S-M) Approach(Elsevier Ltd, 2025-03) Bakir-Güngör, Burcu; Temiz, Mustafa; Canakcimaksutoglu, Beyza; Yousef, MalikColorectal cancer (CRC) is one of the most prevalent forms of cancer globally. The human gut microbiome plays an important role in the development of CRC and serves as a biomarker for early detection and treatment. This research effort focuses on the identification of potential taxonomic biomarkers of CRC using a grouping-based feature selection method. Additionally, this study investigates the effect of incorporating biological domain knowledge into the feature selection process while identifying CRC-associated microorganisms. Conventional feature selection techniques often fail to leverage existing biological knowledge during metagenomic data analysis. To address this gap, we propose taxonomy-based Grouping Scoring Modeling (G-S-M) method that integrates biological domain knowledge into feature grouping and selection. In this study, using metagenomic data related to CRC, classification is performed at three taxonomic levels (genus, family and order). The MetaPhlAn tool is employed to determine the relative abundance values of species in each sample. Comparative performance analyses involve six feature selection methods and four classification algorithms. When experimented on two CRC associated metagenomics datasets, the highest performance metric, yielding an AUC of 0.90, is observed at the genus taxonomic level. At this level, 7 out of top 10 groups (Parvimonas, Peptostreptococcus, Fusobacterium, Gemella, Streptococcus, Porphyromonas and Solobacterium) were commonly identified for both datasets. Moreover, the identified microorganisms at genus, family, and order levels are thoroughly discussed via refering to CRC-related metagenomic literature. This study not only contributes to our understanding of CRC development, but also highlights the applicability of taxonomy-based G-S-M method in tackling various diseases. © 2025 Elsevier B.V., All rights reserved.Conference Object Leveraging MicroRNA-Gene Associations With Mirgedinet: An Intelligent Approach for Enhanced Classification of Breast Cancer Molecular Subtypes(Springer International Publishing AG, 2025) Qumsiyeh, Emma; Bakir-Gungor, Burcu; Yousef, MalikUnderstanding the molecular subtypes of breast cancer is crucial for advancing targeted therapies and precision medicine. For the BRCA molecular subtype prediction problem, this study employs miRGediNET, a machinelearning approach that integrates data from miRTarBase, DisGeNET, and HMDD databases to investigate shared gene associations between microRNA (miRNA) activity and disease mechanisms. Using the BRCA LumAB_Her2Basal dataset, we evaluate miRGediNET's performance against traditional feature selection methods, including CMIM, mRmR, Information Gain (IG), SelectKBest (SKB), Fast Correlation-Based Filter (FCBF), and XGBoost (XGB). These feature selection techniques were assessed using various classification algorithms including Random Forest (RF), Support Vector Machine (SVM), LogitBoost, Decision Tree, and AdaBoost, all executed with default parameters. The feature selection methods were tested using Monte Carlo Cross-Validation, where performance metrics obtained for each iteration were averaged to ensure robustness. Our findings reveal that miRGediNET outperforms traditional methods in accuracy and Area Under the Curve (AUC), emphasizing its superior capability to identify key genes that bridge miRNA interactions and breast cancer mechanisms. Notably, both miRGediNET and Information Gain (IG) feature selection consistently identified ESR1, a critical biomarker frequently reported in recent research associated with breast cancer prognosis and resistance to endocrine therapies. This integrative approach provides deeper biological insights into miRNA-disease interactions, paving the way for enhanced patient stratification, biomarker discovery, and personalized medicine strategies. The miRGediNET tool, developed on the KNIME platform, offers a practical resource for further exploration in the field of bioinformatics and oncology.Article Citation - WoS: 7Citation - Scopus: 7Investigating Strain Rate Effects on Damage Mechanisms in Hybrid Laminated Composites Using Acoustic Emission(Elsevier Sci Ltd, 2025-12) Gulsen, Abdulkadir; Kolukisa, Burak; Etcil, Mustafa; Caliskan, Umut; Zafar, Hafiz Muhammad Numan; Demirbas, Munise Didem; Bakir-Gungor, BurcuHybrid composites, which combine distinct fiber types such as carbon, basalt, and aramid, provide a synergistic balance of strength, stiffness, impact resistance, and energy dissipation, making them appealing for critical applications in aerospace, automotive, and other high-performance industries. Monitoring damage progression in these composites is vital for ensuring structural integrity and preventing catastrophic failures. Acoustic emission (AE) serves as a powerful, noninvasive technique for real-time structural health monitoring, capturing the transient stress waves generated when damage events occur. This study utilizes AE to examine the influence of strain rate on damage modes in carbon/basalt/aramid hybrid composites under three-point bending. An unsupervised feature selection based on Laplacian scores is employed to identify the most relevant AE features with damage modes, while SHapley Additive Explanations (SHAP) are used to evaluate the correlation between AE features and strain rates. The correlation analysis results indicate that peak frequency (PF) serves as a key indicator, demonstrating significant shifts at higher strain rates. Gaussian Mixture Model (GMM) clustering is used to analyze hybrid composites by examining clustered AE signals based on selected features identified through Laplacian scores, with Silhouette scores employed to determine the optimal number of clusters. This study highlights the role of AE in understanding fiber interactions and damage evolution, offering valuable insights into the mechanical performance and optimization of carbon/basalt/aramid hybrid composite structures.Article Citation - WoS: 2Citation - Scopus: 4Engineering Novel Features for Diabetes Complication Prediction Using Synthetic Electronic Health Records(Frontiers Media S.A., 2025-04-14) Voskergian, Daniel; Bakir-Gungor, Burcu; Yousef, MalikDiabetes significantly affects millions of people worldwide, leading to substantial morbidity, disability, and mortality rates. Predicting diabetes-related complications from health records is crucial for early prevention and for the development of effective treatment plans. In order to predict four different complications of diabetes mellitus, i.e., retinopathy, chronic kidney disease, ischemic heart disease, and amputations, this study introduces a novel feature engineering approach. While developing the classification models, we utilize XGBoost feature selection method and various supervised machine learning algorithms, including Random Forest, XGBoost, LogitBoost, AdaBoost, and Decision Tree. These models were trained on synthetic electronic health records (EHR) generated by dual-adversarial autoencoders. These EHRs represent nearly 1 million synthetic patients derived from an authentic cohort of 979,308 individuals with diabetes. The variables considered in the models were the age range accompanied by chronic diseases that occur during patient visits starting from the onset of diabetes. Throughout the experiments, XGBoost and Random Forest demonstrated the best overall prediction performance. The final models, which are tailored to each complication and trained using our feature engineering approach, achieved an accuracy between 69% and 77% and an AUC between 77% and 84% using cross-validation, while the partitioned validation approach yielded an accuracy between 59% and 78% and an AUC between 66% and 85%. These findings imply that the performance of our method surpass the performance of the traditional Bag-of-Features approach, highlighting the effectiveness of our approach in enhancing model accuracy and robustness.