118 No. 60 Alander Extending HP Lattice Model with Non-Local Hydration Jarmo T. Alander Jarmo.Alander@uwasa.fi Department of Information Processing and Production Economics, University of Vaasa, PO Box 700, FIN-65101 Vaasa, Finland Keywords: global optimisation, hydration, lattice model, protein folding, selforganization 1 Introduction It seems obvious that neither a pure random optimisation approach nor a pure deterministic one is able to solve the protein folding problem. A stochastic approach is too slow. Most of the effort is lost in generating hopelessly denatured conformations. A pure deterministic approach is able to avoid most of the bad conformations, but it usually has to process much too many conformations before the native one is found. So many, that too long a time is needed to reach the solution. 2 Method and Results In this work we have attacked the famous protein folding problem by extending a simple HP lattice model [5] by a longer range hydration interaction. The conformation energy optimisation was done using several population based methods: the most fit conformations of a set of randomly generated trial conformations were re-evaluated by slightly modifying them in a Monte Carlo fashion [1,2]. It is proposed that by combining features of both stochastic and deterministic search approaches, we can effectively avoid generating hopelessly bad conformations while not trapping to local extremes. Our approach using in addition a longer range hydration interaction turns out to solve this problem in a way that is not only computationally efficient but also physically consistent [3,4]. The major physical contribution of this work is the modeling and testing of hydration in a lattice model framework. The results indicate that hydration, or similar phenomenon causing longer range interaction, is vital in overcoming the Levinthal’s paradox [6]. More specifically hydration interaction brings the distant parts of a sequence into close association. Hence hydration drives the sequence into more compact, molten globule like, conformations and thus dramatically reduces the otherwise intractable number of possible conformations into a tractable one, out of which the random search by thermal motion is able to reveal the native state. The effect of hydration on conformation search was tested by searching the native conformations of several test sequences (Table 1). As can be seen, in general hydration seems to be beneficial. Without it our algorithm needs considerably more local steps. Similarly it seems to be beneficial to count HH-contacts, but their weight used in the free energy formula does not seem to be so critical. Hence it seems that both local and longer range interactions should be modelled in order to have a realistic fast folding process. Moreover a small population size seems to be better than large. This is easy to explain: the bigger the population size the more time it takes to process the best trials. Population based search is beneficial because search without a population of trials is clearly less efficient than that using a population, however. References [1] Alander, J.T., Protein folding problem, an algorithmic lattice model approach, Licenciate thesis, University of Helsinki, Department of Physics, 1999. ftp.uwasa.fi/cs/report99-1/Main.ps.Z