Loading...

Drug-Target Binding Affinity Prediction Based on Machine Learning Algorithms and Features Extraction using Neural Networks

Belal Yengi Kand, Alireza | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57526 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Koohi, Somayyeh
  7. Abstract:
  8. One of the fundamental steps in drug discovery and development is predicting the binding affinity of the drug to target proteins. Due to the high costs, slow pace, and lack of structural data in experimental approaches, several machine learning and deep learning-based methods have been proposed to date. Machine learning-based methods, which have shown high prediction capabilities, require precise engineering and selection of a wide range of physicochemical, biological, and structural information about proteins and drugs. On the other hand, deep neural network-based methods require the design of numerous complex networks for automatic feature extraction. Therefore, a method that can utilize the capabilities of machine learning algorithms in the prediction phase and the feature-learning capabilities of deep neural networks could provide more accurate predictions with less network complexity. In the proposed method, a model is presented that leverages the feature extraction capabilities of deep networks for proteins and drugs and the predictive power of the gradient boosting machine learning algorithm for predicting drug-protein binding affinity. This method involves two stages: feature extraction and prediction. In the first stage, valuable information about proteins and drugs is extracted using deep neural networks. At this stage, to balance accuracy and network complexity, it is also possible to use descriptor extraction algorithms from proteins and drugs to represent the inputs. In the second stage, the binding affinity is predicted using the gradient boosting machine learning algorithm on the features extracted in the first stage, along with the binding affinity values. According to the schedule, after completing preliminary studies, the desired model will be designed, implemented, and evaluated on various datasets. This model will be compared with past research using accuracy evaluation criteria
  9. Keywords:
  10. Machine Learning ; Deep Neural Networks ; Binding Affinity Prediction ; Target Proteins ; Pharmacy ; Feature Extraction ; Drugs

 Digital Object List

 Bookmark

No TOC