The National Institute of Advanced Industrial Science and Technology (AIST; President: Hiroyuki Yoshikawa) has succeeded in the highly accurate electronic state calculations of a photosynthetic protein complex consisting of more than 20,000 atoms.   The calculation was performed on a cluster-type supercomputer, "AIST Super Cluster," by using the fragment molecular orbital (FMO) method.

The results of this work were presented at the International Conference for High Performance Computing Networking Storage and Analysis held in Seattle in the USA, and received high acclaim because of its advanced computational technique and detailed scientific results. We are the first Japanese researchers to win the best technical paper award of the conference.

AIST has been operating a cluster-type computer, the "AIST Super Cluster," with the highest total computational performance in Japan since May 2004.  The facility provides rich computational resources for nanotechnologies and bioinformatics at a low cost, and is utilized for life science research, e.g., to clarify the functions and behaviors of biological materials such as proteins. We have also developed a novel algorithm, the fragment molecular orbital (FMO) method, to solve a macromolecular electronic state, which is well suited on cluster computers.

By implementing the FMO method on the AIST Super Cluster, we have performed the largest-ever electronic state calculation on the biomolecule.  The results were submitted as a technical paper to the International Conference for High Performance Computing Networking Storage and Analysis (SC|05).  SC|05 is the largest international conference in this field with ten thousand participants, and is known by the high quality papers presented therein.  This year, 63 papers were accepted among 260 (acceptance ratio of 24.2 %), of which our paper has received the best technical award.

The target is a membrane protein complex of the purple photosynthetic bacteria Rodopseudomonas viridis.  The system is composed of four protein chains, in which a group of molecules, called the electron transfer system, is buried.  The electron transfer system is known to play an important role in the photosynthetic reaction, where the energy of light is converted to chemical form.  On the other hand, the functionality of the surrounding proteins is not resolved yet.  To assess the detailed interaction between the electron transfer system and the proteins, the electronic state calculation of the whole system including the proteins is indispensable.  The conventional methods, however, are hard to be applicable because they require enourmous amount of computation.

In this work, the FMO calculation on the photosynthetic system was carried out by using 300 nodes (600 CPUs) on the P-32 subunit of the AIST Super Cluster.  The large-scale parallel calculation finishes in 73 hrs, verifying that our method enables the accurate electronic state calculations of macromolecules with a high cost-performance ratio (Figure 1).

The present achievement is a step further to clarify the functional mechanism of the photosynthetic reaction center. It may be extended, in the future, in designing artificial photochemical systems, and hopefully be applied to the prevention of global warming due to carbon fixation and the solution of the food problem. 

The high acclaim for our work in SC|05 demonstrates that the development not only of high-speed hardware but also of purpose-adequate software is important in supercomputing, and thus the establishment of supercomputer-utilizing technology is desired

Figure 1 Isosurface of an electron density of the photosynthetic protein complex.