![]() | Only 14 pages are availabe for public view |
Abstract Computational techniques are used widely to help in the drug design process for discovering new drugs. The use of computational techniques has led to saving a lot of money and effort for pharmacists and chemists. Docking is one of the computational applications in this area for estimating the binding strength, represented as energy. Docking simulates the protein-ligand interaction process in order to give detailed perspective of what happens in the body. There are many approaches for computational docking to predict the ability of a ligand, or chemical compound, to inhibit a specific protein. Machine learning techniques are expected to help in this prediction when applied to available experimental data. In this thesis, we worked on two directions. First, we used computational docking to study the binding between a drug, and a group of cancer-related proteins to investigate the possibility of suggesting an untraditional cancer treatment. Second, we proposed and followed two methodologies to use and evaluate the ability of machine-learning techniques to predict binding and inhibitory activity of ligands against specific proteins. This prediction is based on available experimental dataset, and the calculated chemical descriptors of compounds. The prediction was done in two ways, classification and regression with feature selection |