Exploratory visualization of biological networks

Understanding the complex relationships between biological entities has a great significance in drug discovery and development. There has been a substantial investment of research effort for this purpose. However, attempts for drug discovery and development are time consuming and costly. Thus, researchers seek computational approaches to reduce the time and cost and minimize the risk of adverse events of clinical experiments. In recent years, advancements in biomedical research amplify the vast accumulation of biological data. Consequently, the known relationship and attribute information of biological entities can be used to model biological relationships in the manner of biological networks to understand the complex relationship between biological entities along with revealing latent relationships that can excel the discovery of potential therapeutic treatments. In this thesis, we introduce a dimensionality reduction-based framework, named I-UMAP, to model complex biological relationships. The proposed framework accurately models the biological networks, in our case drug--target interaction networks, to be used in exploratory data analysis to comprehend the essential biological relationships. Since the introduced framework exploits the known relationship information along with entity properties, the crucial interactions between biological entities can also be unveiled. Therefore, this task can be formulated as a link prediction problem amongst biological entities. We introduced a novel approach to benefit from both the known interaction information and entity features. In order to enhance the capacity of the base algorithm UMAP, we modified the UMAP algorithm to improve the optimization performance. To test the efficacy of our algorithm, we conducted experiments on drug--target interaction networks and drug--disease interaction networks. The results of the conducted experiments demonstrated that I-UMAP surpasses the performance of UMAP algorithm and could be beneficial to explore and unveil biological relationships.