Multilevel Parallelization of Software for Accurate Protein-Ligand Affinities

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Many drugs are small molecules that act by binding to a specific protein and thus blocking or altering its actions. For example, the HIV protease inhibitors are important AIDS treatments that work by binding in the active site of the protease enzyme and preventing it from helping to make new viruses. When scientists identify a protein, like HIV protease, as being important in a disease process, a next step often is to determine its three-dimensional structure in great detail. This structure then provides valuable guidance to chemists trying to design a small molecule that will bind the protein tightly. However, even when they know the structure of the protein, there is still a lot of trial and error in designing a drug. Many researchers have worked on computer programs to help predict whether a given molecule will bind a given protein, but without much success. Now, new software that VeraChem has been developing over the last few years is giving very good results for this problem. However, the software takes a long time to run and would be far more useful if it were much faster. For example, if chemists had an idea for a new compound to try, they could get the answer in a minutes instead of a few days. They could use the method to quickly and cheaply test thousands of compounds in chemical catalogs. And they could check whether a compound that works against their protein would keep working against mutant forms of the protein and thereby avoid drug-resistance. Thus, a fast version of VM2 would be very useful and would be a valuable commercial product. Speeding up VeraChem's method, VM2, is not as simple as running it on a faster computer, because individual computers have not been getting much faster in recent years. What is changing, though, is that computers are being made with more and more processors. The goal of this project is to speed up VM2 enormously by spreading its computational work across large numbers of separate processors in a single computer, in a cluster of computers, and even in a video card. This is not a simple task, but researchers have been able to speed up related molecular calculations in this way, and we are confident the same can be done for VM2. PUBLIC HEALTH RELEVANCE: We want to let scientists design new medicines more quickly with a computer program. The problem is that the program takes too long to do its calculations. This project is to speed up the calculations by changing the program so that it can make a large number of computer processors to work together to calculate the answers in a short time.