Supercomputer targets cellulose bottleneck

Scientists are using ‘virtual molecules’ modelled on a supercomputer to investigate how enzymes break down cellulose, with the aim of speeding up the process.

Producing ethanol from cellulose is slow and expensive, the central bottleneck being the sluggish rate at which the cellulose enzyme complex breaks down tightly bound cellulose into sugars.

To help unlock the cellulose bottleneck, a team of scientists has conducted molecular simulations at the San Diego Supercomputer Centre (SDSC), based at UC San Diego. By using ‘virtual molecules,’ they have discovered key steps in the intricate dance in which the enzyme acts as a molecular machine. They attach to bundles of cellulose, pull up a single strand of sugar, and put it onto a molecular ‘conveyor belt’ where it is chopped into smaller sugar pieces.

In their simulations, the scientists found that initially the binding part of the enzyme moves freely and randomly across the cellulose surface, searching for a broken cellulose chain. When it encounters an available chain, the cellulose itself seems to prompt a change in the shape of the enzyme complex so that it can straddle the broken end of the cellulose chain. This gives the enzyme a crucial foothold to begin the process of digesting or ‘unzipping’ the cellulose into sugar molecules.

In real life the process occurs far too quickly to evaluate visually, but by using the supercomputer simulations to break the events into a step-by-step process, the scientists can see the precise details of the role of velocity, trajectory, movement, and arm angle. To undertake the large-scale simulations, the researchers used the CHARMM (Chemistry at HARvard Molecular Mechanics) suite of modelling software.

According to the researchers, an accurate understanding of the key molecular events required the simulations to run for some six million time steps over 12 nanoseconds in order to capture enough of the motion and shape changes of the enzyme as it interacted with the cellulose surface.

This is an extremely long time in molecular terms, and the computation-hungry simulations ran for some 80,000 processor-hours running on SDSC’s DataStar supercomputer.