Computer-aided Drug Screening And Design On Avian Influenza Virus Neuraminidase Inhibitors

The discovery and development process of the traditional anti-avian influenza drug was time-consuming, laborious and costly as it was screened from a large quantity of highly active compounds without any direction. Computer-Aided Drug Design(CADD)is based on computational chemistry. It can be used to design and optimize the lead compounds after figuring out the relationship between the biological macromolecules and the ligands through computer simulations and computations. It not only makes a rapid progress in the study of new drug discovery but saves lots of money and time and it has been an important method in new drug discovery. This paper uses the methods of CADD to screen and design the lead compounds to guide the anti- avian influenza drug design. It can provide a directional guide in the process of drug discovery and save lots of labors. The main content is as follows: Firstly, based on the "lock and key model", Semi-flexible docking were implemented on the crystal structure of neuraminidase which is downloaded from the Brookhaven Protein Data Bank (PDB entry: 2QWK) and the 126 small-molecule inhibitors built by Sybyl7.0 packages. Then these models were optimized by genetic algorithms and simulated annealing, followed by semi-empirical scoring evaluation.Through the docking process, we got the optimal conformations of the inhibitors and their bonding models, and constructed the linear relationship between the lowest bonding energies and the corresponding activities .Through the test. We found the linear model highly reliable and predictable. After analyzing the bonding models and the structural features of small-molecules, we found that the bonding models, the small-molecule structures and the activity values are fully consistent with each other, which once again proved that the linear model was reliable.Secondly, a database of small molecules with their optimal conformations was built. According to the rules of composite and the molecular structure ,the inhibitors were aligned and the three-dimensional quantitative structure-activity relationship (3D-QSAR) of both the divided and overall group were built ,including the comparative molecular field analysis model (COMFA) and the comparative molecular similarity indices analysis (COMSIA) model. Finally, the relationship between small-molecule inhibitory activity and characteristics of the molecular field was built by the partial least squares (PLS). Moreover, the molecular potential energy surface maps were also given to make the explanation of the model easy and intuitive.Thirdly, the model was tested on the data of the test set. It turned out that both the divided group and the overall model have advantages, that is, the divided group model has a higher prediction capability while the overall model has a wider application. We will get more reliable predictions on the specific test if combination of the two methods was used. By comparing the molecular potential energy surface maps with the bonding model, we found that they were fully consistent with each other, which once again proved that the 3D-QSAR model is reliable.