Book of Abstracts: Albany 2011

category image Albany 2011
Conversation 17
June 14-18 2011
©Adenine Press (2010)

Swarm intelligence: Cooperative replica methods for prediction of protein structure

The use of atomistic simulation techniques to directly resolve protein tertiary structure from primary amino acid sequence is hindered by the rough topology of the protein free energy surface and the resulting simulation timescales required. An interesting approach to improve the exploration of this surface, termed the SWARM-MD method, was proposed by Huber and van Gunsteren (1). Their approach is inspired by particle swarm methods. Introduced in 1995 and originally applied to electrical circuit design, the particle swarm algorithm mimics the social behaviour of bird flocking or fish schooling. The efficiency of the algorithm’s search behaviour, which exhibits remarkable organization and planning, can be traced to cooperativity of individual members of the swarm, influenced by their memory and that of their peers. In SWARM-MD, the free energy surface is smoothed through the use of a swarm of multiple interacting simulation replicas that are driven towards the average conformation of the swarm members. We have successfully applied this approach to prediction of the native states of a series of model peptides (2), including Trp-cage miniprotein in aqueous solvent. In each case, the cooperation between replicas was found to improve the convergence of the simulations towards the native state. Limitations and future directions of this method will be discussed. This research has been supported by the EPSRC.

  1. T. Huber and W.F. van Gunsteren, J Phys Chem A 102, 5937-5943 (1998).
  2. N.J. Bruce and R.A. Bryce, J Chem Theory Comput 6, 1925-1930 (2010).

Neil J. Bruce
and Richard A. Bryce*

School of Pharmacy and Pharmaceutical Science, University of Manchester, Manchester, M13 9PT, UK

Ph: +44-161-275-8345
Fx: +44-161-275-2396