By employing a novel computational strategy, researchers have mapped the electrical landscape of biological molecules made up of more than 1 million atoms. Previous methods were typically limited to fewer than 50,000 atoms.
Electrostatic properties play an important role in the stability and dynamics of proteins, nucleic acids, and other biomolecules. With the new approach, scientists can model electrostatic interactions in functional parts of cells. These include microtubules, which usher nutrients and other substances back and forth within the cell, and ribosomes, which serve as protein-making centers.
“This work signals a new era of calculations on cellular-scale structures in biology,” says chemist J. Andrew McCammon of the University of California, San Diego (UCSD) in La Jolla. McCammon, Nathan A. Baker of UCSD, and their coworkers report their findings in the Aug. 28 Proceedings of the National Academy of Sciences.
To model how the charges of individual atoms interact to produce a molecule’s electrostatic potential, or field, researchers solve the so-called Poisson-Boltzmann equation for points throughout the molecule. They use the equation to calculate the potential at each point on a three-dimensional grid within a box enclosing the molecule.
Last year, UCSD mathematicians Michael J. Holst and Randolph E. Bank demonstrated that it’s possible to use a quick, rough solution for widely separated points to guide detailed calculations in much smaller regions.
Baker and Holst adapted the strategy for electrostatic modeling of biomolecules.
Their approach parcels out the computation to a large number of processors in a novel way. Each processor, either in a single supercomputer or across a network of computers, solves the equation for the same grid of widely spaced points, then focuses on a different tiny piece of the molecule. The processor uses the first, rough solution as a guide to arrive at the second, highly precise solution. A master processor then assembles the individual results into a detailed electrostatic portrait of the biomolecule.
The research team tested the method on a 1.25-million-atom microtubule. In less than an hour, 686 processors in the IBM Blue Horizon supercomputer at the San Diego Supercomputer Center produced an electrical map of the structure. No computer could have done the calculation in a practical amount of time using previous methods, the team notes.
One immediate goal, Baker says, is to “get the software distributed so a wider audience can start using it to perform calculations . . . on large molecules of interest.”
