Faculty of Computer Science and Engineering
Permanent URI for this communityhttps://repository.ukim.mk/handle/20.500.12188/5
The Faculty of Computer Science and Engineering (FCSE) within UKIM is the largest and most prestigious faculty in the field of computer science and technologies in Macedonia, and among the largest
faculties in that field in the region.
The FCSE teaching staff consists of 50 professors and 30 associates. These include many “best in field” personnel, such as the most referenced scientists in Macedonia and the most influential professors in the ICT industry in the Republic of Macedonia.
Browse
4 results
Search Results
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Graph and Convolutional Methods for Advancing Ligand Binding Affinity Modeling in Drug Research(IEEE, 2025-06-02); Drug discovery increasingly relies on accurate ligand binding affinity modeling to reduce the cost and time spent on trial-and-error experiments. However, existing computational methods often exhibit limited generalization, interpretability, and training efficiency. To address these gaps, we present a novel framework that integrates graph neural networks (GNNs) and convolutional models to model proteinligand interactions. Our approach builds on recent findings that highlight the benefits of representing protein-ligand complexes through graph topologies while capturing spatial and structural features using convolutional layers. We examine two publicly available datasets, PDBbind and BindingDB, both of which exemplify diverse protein-ligand complexes. Comprehensive experiments demonstrate that our integrated GNN-convolutional model improves predictive accuracy, reduces computational costs, and enhances interpretability. In addition, ablation studies reveal the roles of graph structural encoding and convolutional feature extraction in capturing crucial interaction signals. Theoretically, our study augments existing research by providing empirical evidence that unifying graph and convolutional strategies can enrich the insight into topological and spatial representation learning in ligand binding affinity prediction. Practically, the proposed framework can be readily adopted in workflows where large-scale exploration of protein-ligand complexes is required, potentially accelerating early-stage drug discovery by refining virtual screening and lead optimization. This work closes previously identified performance and interpretability gaps, offering a rigorous pathway to future applications in ligand binding affinity modeling. - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Bridging Gaps in Ligand Binding Affinity Prediction: Empirical Machine Learning Analyses(IEEE, 2025-06-02); Predicting ligand binding affinity remains a critical challenge in computational drug discovery, as existing techniques often require extensive computational resources and are not readily generalizable to diverse protein-ligand systems. This study addresses three key gaps in current research: (1) the lack of versatile, data-efficient predictive models; (2) insufficient strategies for integrating protein and ligand structural information; and (3) limited methods for simultaneously improving accuracy and generalizability. By systematically reviewing recent advances in machine learning-including approaches derived from deep learning, graph-based methodologies, and hybrid frameworks-we show how emerging techniques enable higher predictive power and reduced computational cost. We leverage two large-scale, public datasets (PDBBind and BindingDB) to empirically evaluate a novel dual-model framework that integrates graph-based feature extraction and neural network regression. Comparative analyses illustrate how spatial and sequence representations contribute to model performance, achieving robust improvements in binding affinity prediction. The theory-based advantages of this approach demonstrate how it reveals both small-scale and wide-ranging relationships between proteins and their ligands and the operational benefits result in quicker medication development through decreased processing needs. The research confirms the necessity of building adaptable frameworks which unite structural data with sequence data for therapeutic advancement along clear research paths. - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Assessing Personalized Engineering Learning Experience with a Multi-Modal AI Tutoring Framework(IEEE, 2025-06-02); Fetaji, FjollaThis paper presents an innovative and novel multimodal AI tutoring framework in the effort to increase the effectiveness of personalized engineering learning experiences. Filling the gaps in adaptive tutoring systems and considering people's emotional engagement, the framework combines the cognitive load theory, indicators of emotional intelligence, and adaptive learning algorithms to develop an overarching, context-specific instructional landscape. The study makes use of a mixed-methods design, using machine learning driven tutoring interfaces, state of the art learning analytics, and sentiment analysis on a long-term study with 68 engineering students. Incorporating powerful affective computing techniques and intelligent intervention methods, this study presents an encompassing approach that can tune the scaffolding support to learners' abilities and adjust to everchanging competencies of learners while being proactive about detecting the indicators of cognitive overload. The uniqueness aspect of this research lies in its synergistic combination of various data streams (multimodal) - text, visual, and biometric data - combined with dynamic AI-based recommendation model, which maximizes personalized feedback loops. Theoretically, the study expands the insights into how the cognitive load and the emotional dynamics form learning outcomes. Practically, it provides a scalable, flexible approach to the integration of multimodal AI tutors in various educational environments. This work helps to bridge the gap between the cognitive science principles and educational technology solutions and offers new findings regarding designing effective, user-centric intelligent tutoring systems that increase knowledge retention, create motivation, and increase the engineering education outcomes. - Some of the metrics are blocked by yourconsent settings
Item type:Publication, A Survey of Graph Neural Network Architectures in Ligand Binding Affinity Prediction Models(IEEE, 2024-05-20); Ligand affinity prediction plays a pivotal role in drug discovery, influencing the efficiency and success of drug development processes. Traditional methods struggle in accurately capturing the complex interactions within molecular structures, prompting the exploration of advanced techniques such as Graph Neural Networks (GNNs). This paper provides an analysis of GNNs in the context of ligand affinity prediction, exploring their architecture, applications, and potential impact on revolutionizing drug discovery. Our findings suggest that GNNs can offer improvements over traditional computational methods, particularly in handling the dynamic and complex nature of molecular interactions. We highlight innovative GNN architectures that have shown notable success in predicting ligand binding affinities, such as heterogeneous graph representation and attention mechanisms. The implications of these advancements suggest a paradigm shift in drug discovery, where GNNs can lead to more accurate predictions and accelerate the identification of potential drug candidates. This study underscores the transformative potential of GNNs in enhancing predictive accuracy and efficiency in drug development.