Arieh Warshel, from the University of Southern California, U.S., is one of the scientists to win the Nobel Prize in Chemistry for 2013. His research group’s homepage has computer simulations of molecular interactions with an apt caption: ‘Stuff that you won’t find at Netflix.’ One simulation depicts how the gene RAS functions as a molecular switch inside a cell in a two-minute film.
There are many aspects to this simulation that are mind-boggling. These interactions are simulated at molecular levels, dealing with distances on the nanometre scale. Simpler classical physics simulations cannot yield agreeable results at this scale and require the use of quantum physics techniques. Computer programs and algorithms developed in the last three decades are making realistic molecular simulations possible, which, in turn, are used to make drugs to cure chronic diseases such as cancer.
When a drug is administered into the human body to fight a disease, it works by either inhibiting or accelerating the function of special proteins in cells called enzymes. To discover newer drugs, it is imperative to understand how a drug would interact with an enzyme. Before the advent of computer simulations this was usually achieved by testing them on animals or humans. The major portion of the work involved in simulation is to observe the shape an enzyme would take. When inactive, enzymes are long strands of proteins. They become active when they fold into more stable, three-dimensional shapes.
In testing vehicle safety, it is not necessary to conduct all the tests with real cars bumping into various obstacles. Reliable — and perhaps even better — results can be obtained by modelling cars and the environment using sophisticated computer programs.
The reliability of these tests, nonetheless, depends on how many parameters are taken into consideration, and to what precision the simulation is carried out, which, in turn, crucially depends on the computational resources available for the experiment.
Similarly, to understand how drugs interact with the human body, not all drugs need to be tested on animals or humans. The molecular behaviour of the drug can give deep insights into the kind of impact the drugs may have on the observed cells. The simulation of drug-enzyme interaction and the observation of the protein folding process require enormous computational power. This is because the interplay of the hundreds of parameters requires complex modelling techniques.
The results from a computer simulation using classical equations in physics are mostly deterministic. That is, for a given set of definite inputs, even sophisticated computation would usually yield a certain specific result. However, modelling systems based on quantum physics turns them into probabilistic systems. The computation in such a system is not only much bigger but involves a wider range of computations. In these biochemical simulations, a combination of both the approaches is required, which results in a significant increase in the computational burden, implying more processor cores, extra physical memory, greater storage space and intelligent programming.
The high-resource demand in biochemistry simulations is manifested in the Folding@home project conducted by scientists at Stanford University. On the homepage of this project, visitors are urged to contribute their computer resources by sharing the load of protein folding computations using distributed computing.
About 3,00,000 users across the world are contributing their computing resources by allowing a computer program to run in the background on their systems, forming the world’s largest distributed supercomputer.
When a volunteer opts to run these nodes, servers at the Folding@home project assign smaller tasks to the client, which are monitored by the control application installed on the volunteers’ computers.
When a task is completed within a deadline, the volunteer is awarded ‘credit points’ and the next task is carried out. Folding@home servers orchestrate this parallel computation and harness the computational power of all volunteer computers.
The number of computations in simulating protein folding is way beyond the capabilities of even a very powerful computer. Before proteins start folding, they are long strands of polypeptides. Imagine these to be really long strands of beads attempting to form a stable 3D structure. Every node has a maximum of 6 degrees of freedom (forward/back, up/down, left/right, pitch, yaw, roll), and when simulating the overall structure, it is important to compute the shape this strand might take with increments in each of the degrees of freedom, at every node and observe the pattern that would yield the highest stability.
Using distributed computing, researchers at Stanford want to simulate as many possible protein foldings as possible, and in the least time possible. Some of the simulations can result in drugs that can help cure diseases such as cancer or fight HIV.
As a result of work done since the 1980s, researchers now have desktop tools to visualise 3D structures of biological molecules. One widely used free and open source tool based on Java is Jmol. This tool is a 3D visualiser of chemical structures with features for chemicals, crystals, materials and biomolecules.
As the number of processor cores stacked in a processor increase, parallel computing with networked and distributed computing, and agile software platforms have reduced the time taken to perform high-end simulations. This may appear to be a small step in the evolution of computational capabilities, but they are a giant step for humanity because they bring the hope of cures for diseases such as Alzheimer’s and cancer, caused by protein misfolding, closer to being realised.