After deciding to extend and deepen my knowledge about bioinformatics, I joined the structural bioinformatics group of Torsten Schwede at the Swiss Institute of Bioinformatics. So far, I was interested in better understanding how drug-activated transcription factors bind to the DNA of our genome, now, I turned to the interaction between drugs (or generally: small chemical molecules) and proteins themselves. This problem is also known as protein-ligand docking, and aims to predict for any given small chemical molecule in what three-dimensional orientation and with what strength it would bind to a particular protein. This has applications not only in the prediction of adverse drug effects, but perhaps more importantly in the discovery of new drug candidates.

We are working on Dengue fever, a neglected viral disease that affects approximately 100 million people every year, and which is transmitted from human to human by the mosquito Aedes aegyptii. There is no vaccine or specific drug treatment available, and the disease is a world-wide problem with more than 2 billion people living in affected regions, and the area where the mosquito . Scarily, dengue infections seem to be on a sharp incline lately (article in German), and no-one understands yet why.

Active site of the Dengue NS5 methyltransferase protein:

In our project, we have identified suitable target proteins among those produced by the virus, and are now trying to find chemical molecules that potentially inhibit these enzymes. If we manage to block the virus from infecting human cells and replicating in there, we have a compound in our hands that can be developed further, and hopefully one day become a drug against Dengue. However, we are well aware that in our simulations, we cannot compute a drug. We’re therefore collaborating with scientists that take our suggestions and test them in the laboratory and in cell culture labs. More information about this project can also be found here.

Ligand docking simulations need a lot of computing power. For millions of molecules, a search has to be performed that tries to estimate the best-finding pose of the molecule, and the binding energy associated with it. These computations can’t be performed on a single desktop PC, and would even block computer clusters for long periods of time. Therefore, I have turned to grid computing as a way of finding sufficient computing power.

No Comments »