Protein structure prediction from sequence alone by "brute force" is a computationally expensive problem. Feldman and Hogue have developed the fastest known method for generating all-atom off-lattice samples of protein conformational space. Howard Feldman discusses the methods for solving the two halves of the protein folding problem - sampling and scoring - and how distributed computing can be used to rapidly sample vast areas of conformational space and improve the success at prediction, as described in the paper:

Feldman,H.J. and Hogue,C.W.V. Probabilistic Sampling of Protein Conformations: New Hope for Brute Force? Proteins. (2002) 46:8-23.