Background

This research focuses on providing accurate computational methods for better understanding the 3-Dimensional properties of protein-ligand interactions; this is important in the development of pharmaceuticals. The traditional approach of wet-lab experiments can be expensive in terms of people, money, and resources. Thus computational methods, such as 'molecular docking', are being used to support wet-lab experiments.

The key in understanding a certain protein-ligand interaction lies in understanding the struture and dynamics of the end-product of the interaction.

At GCLab, in order to run the molecular docking simulations which are a key component of the Docking@Home project , CHARMM (Chemistry at Harvard Macromolecular Mechanics) is used. CHARMM is a powerful tool used to study the structure and dynamics of macromolecules.

A docking attempt in Docking@Home, given a protein and ligand, consists of hundreds of thousands of trials and for each trial a random conformation of the ligand is generated. For each generated ligand conformation, several ligand orientations are randomly generated and docked into the protein. Then a molecular dynamics simulation is undertaken which consists of a heating phase followed by a cooling phase. Afterwards, the simulation is refined by running an energy minimization. Finally, the lowest energy structure is returned as the solution for the docking problem. The lowest energy structure is found by ranking the docked ligands with a scoring function which uses the free energy of binding and potential energy.

In the search for the lowest energy structure several other parameters can be considered, other than number of random ligand conformations and orientations like force-field acting on the atoms of the complex and the protein representation.

The purpose of this project is to perform an analysis of the docking results in order to determine whether there is a correlation between the force field used in the dockings and the accuracy obtained. Twelve force fields were considered over 33 protein-ligand complexes.

This was attempted previously, but no conclusions were made due to the small number of samples collected in the preliminary work.

My Task

My task for the summer is to systematically run a larger number of docking attempts for the different force fields and lattice sizes. After which I will analyze the statistical relevance of the findings and cluster the ligands in accordance to their common behavior. As a final step, I will extract common characteristics (3D features) of the ligands belonging to the same cluster and select the force fields that maximize the inter-cluster classification.

The results of my research will aid in designing and implementing more efficient methods to automatically classify ligands in clusters.