Media of University of Babylon | PhD Dissertation at the College of Information Technology Discusses Gene Regulatory Network Analysis and Prediction Using Graph Neural Networks

Media of University of Babylon - كلية الطب

PhD Dissertation at the College of Information Technology Discusses Gene Regulatory Network Analysis and Prediction Using Graph Neural Networks

by : Duhaa Fadill Abbas

Date : 11/6/2026 51:03:11 PM

Veiw : 97

PhD Dissertation at the College of Information Technology Discusses Gene Regulatory Network Analysis and Prediction Using Graph Neural Networks

Duhaa Fadill Abbas

The Software Department at the College of Information Technology, University of Babylon, witnessed the defense of a PhD dissertation presented by Sura Ibrahim Mohammed Ali Aziz, entitled “Analysis and Prediction of Gene Regulatory Networks (GRNs) Using Graph Neural Networks.” The dissertation was supervised by Assistant Professor Dr. Suraa Zaki Naji and was defended at 9:00 a.m. on Thursday, June 11, 2026, in an academically rigorous environment that reflected the growing importance of employing artificial intelligence and deep learning techniques to address complex biological challenges and support interdisciplinary research that integrates computer science with life sciences.

The dissertation focused on Gene Regulatory Networks (GRNs) as one of the most important computational frameworks for understanding regulatory interactions among genes and their influence on complex biological processes. It highlighted the major challenges associated with constructing GRNs using deep learning techniques, particularly the scarcity of true regulatory links and the highly sparse nature of biological network data. These challenges make it difficult to distinguish meaningful biological interactions from noise, thereby limiting the predictive performance and generalization capability of existing models.

To address these limitations, the researcher proposed a novel hybrid framework named CGDCA-DiGAT, which integrates multiple sources of information to enhance robustness and improve predictive performance in sparse gene regulatory networks. The proposed framework consists of three interconnected stages. The first stage employs GraphSAGE combined with community-based features to predict gene–gene interactions. The second stage utilizes Graph Attention Networks (GAT) enhanced with random-walk representations to predict direct transcription factor–gene interactions. The third stage integrates the outputs of the previous stages to infer indirect regulatory relationships and construct a more accurate and comprehensive gene regulatory network.

The dissertation aimed to uncover hidden regulatory relationships among genes, improve the reconstruction accuracy of GRNs under limited-data conditions, and enhance the prediction of biologically significant regulatory links that have not been previously observed. Experimental evaluations conducted on E. coli and S. cerevisiae datasets demonstrated that the proposed model outperformed several state-of-the-art approaches in the field. The framework achieved an AUC score of 96% on the E. coli dataset and 87% on the S. cerevisiae dataset, confirming its effectiveness and reliability in analyzing complex gene regulatory networks and improving the accuracy of regulatory relationship prediction.

This dissertation reflects the Software Department’s commitment to supporting advanced scientific research that leverages artificial intelligence and large-scale data analytics to address contemporary scientific challenges. It also contributes to strengthening the integration of computer science and biological sciences while opening new avenues for research and innovation in biomedical and life science applications.