Digital science innovation: How computers help to develop health care treatments.
A story about how one makes a new drug
By Anne-Sophie Schillinger
09/29/2015
Making a drug is not easy. It is a long process that can take up to 20 years – or more. Often, drugs have side effects and treat the symptoms of a disease rather than the cause. Why is that? The lack of knowledge, about how the body works at the molecular level, is directly preventing the design of new medicines that could effectively treat diseases with less side effects. Molecular information can be quite complex and difficult to interpret. Often, this relates to the analysis of “big data”. This is where digital science comes into play and where we use computer aided systems to tackle these problems. Technological progress in the past decades has made computers powerful enough to handle biological information and treat biological data more effectively than ever before. So how does one make a new drug with a computer?
Let me first walk you through the basic steps of what a drug is and how one designs it. Many of you may know aspirin, that it has been used since antiquity and was discovered in a plant, the willow tree. In fact, aspirin is a small chemical molecule that directly targets and docks to an enzyme present in the human body, subsequently relieving inflammation. This ancestral remedy is now one the most widely used (and known) drugs, but it does not quite work for everything. There are two ways to discover a new drug that I will mention in this article. The first is to test known drugs and do screening of millions of compounds against a biological target, usually an enzyme. This brute-force technique has enabled the discovery of many new drugs, but is time consuming for use on all identified biological targets. The other way to discover a new drug is to engineer it, i.e. to design it at the chemical level. In order to do this, one needs very precise knowledge of a particular biological target. Scientists needs to know the 3D structures of an enzyme, a “map”, that would essentially state how big the molecule is and how it is organised at the atomic level.
Enzymes are biological molecules that are relatively “big” and are made of thousands of atoms. This complexity is difficult to imagine from the top of your head, which is why we use computers – a lot them. By feeding the computer with the chemical structure of a target of therapeutic interest, i.e. an enzyme, we can get information about what it looks like and where it is possible to dock a drug that will affect its function, and design a drug accordingly. Sounds simple, doesn’t it? Let me add a layer to this. Very often, a biological molecule will slightly change its shape through time, either upon interaction with another biomolecule and/or to perform its function. The behaviour (the movements) of the molecule through time should then be taken into considerations too. Simulations using supercomputers can be done to get this information. Hundreds of processors are used in parallel to perform these simulations. The “behaviour” of the molecule through time will then be computed using its 3D structure, together with Newton’s equation of motion and integration algorithms. This requires millions and millions of calculations and reflects the complexity of biological information to be treated. However, once the simulations are done, one is able to design a drug in a very specific manner: what chemical properties shall that drug have, how big can it be, how can the docking of the drug be optimised to one particular target. Optimisation is very important, because many biological molecules look alike. The specificity of the drug minimises its potential side effects and prevents it to act on similar molecules. This is what is called early stage drug development, where molecular dynamics simulations are being used, as experts would name them.
The computational biology unit at the university of Bergen implements these types of simulation and conducts studies aimed at building knowledge on how the immune system is regulated in some types of inflammation disorders, namely arthritis. So far, simulations using supercomputers have been used to investigate an enzyme, Proteinase 3, as an identified target in this disease, and elucidate its behaviour with relation to the membrane of white blood cells. Enzymes do not work alone and often simulations include other biological entities, like other enzymes, DNA or in some cases models representing a cellular membrane. With this knowledge, drugs can be designed to target pertinent sites of an enzyme. This new era of knowledge building through the computation of biological data was only made possible recently with the increase of computing capabilities, contributing tremendously to biomedical research. Now you know the story of how digital science can help to develop health care treatments.