Loading...

An Efficient Solution for Drug Target Interaction and Binding Affinity Prediction Using Deep Learning Methods

Kalemati, Mahmood | 2024

0 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 57764 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Koohi, Somayyeh
  7. Abstract:
  8. Predicting the interaction and binding affinity of drug-target represents a crucial yet complex phase in the time-consuming and costly process of drug discovery and development. Advances in deep learning have significantly enhanced the ability to model and extract intricate relationships and patterns from diverse biological and pharmaceutical data. However, existing methodologies encounter several fundamental challenges, including the modeling of protein and drug representations, understanding molecular interactions, and overcoming data access limitations. Addressing these challenges has necessitated the use of various heterogeneous architectures, methods, and data structures. Despite these efforts, achieving optimal accuracy remains elusive, often resulting in prolonged model training and testing processes. Therefore, there is a pressing need for efficient solutions to address these challenges and improve the prediction of drug-target binding affinity using deep learning techniques. This research identifies and analyzes the core challenges associated with predicting drug-target binding affinity, proposing innovative solutions. For protein representation, we introduce a novel approach involving orthogonal application of evolutionary and compression-based features in information theory to optimize protein sequence encoding. For drug representation, we present a model based on inception networks designed to extract multi-scale features. Addressing the data access challenge, we propose a model leveraging convolutional generative adversarial networks and a transfer learning mechanism. Modeling molecular interactions in the binding region is approached through the extraction of hierarchical features using capsule networks and dynamic routing, providing an interpretable model for enhanced network understanding. The efficacy of these models is evaluated using rigorous evaluation criteria and diverse datasets. Our solutions not only improve prediction accuracy compared to previous methods but also introduce innovations in network structure and data processing, simplifying and accelerating the prediction process. This highlights the potential of our approach in accelerating drug repurposing efforts, facilitating novel drug discovery, and ultimately enhancing disease treatment
  9. Keywords:
  10. Binding Affinity ; Binding Affinity Prediction ; Interactions ; Proteins ; Deep Learning ; Pharmaceutical Compound ; Drug-Target Interaction

 Digital Object List

 Bookmark

No TOC