Fragment-based Protein Folding Simulations

My lab has pioneered a number of methods for predicting the native folded conformation of a protein from its amino acid sequence. Here are some examples of our successful work in the area of ab initio protein structure prediction.

This animated GIF image shows a synthetic folding trajectory for a small alpha-helical protein (porcine NK-lysin) which was predicted using our FRAGFOLD software as part of the 2nd CASP experiment carried out in 1996. This was the first successful prediction of a novel protein fold in CASP and was also notable as being the first fold prediction method to employ fragment assembly. The basic idea of fragment assembly is to take fragments of already known protein structures and to recombine them randomly to create possible new protein folds. See the following reference for more details:

Jones, D.T. (1997) Successful ab initio prediction of the tertiary structure of NK-Lysin using multiple sequences and recognized supersecondary structural motifs. PROTEINS. Suppl. 1, 185-191.

Each frame of the animation was generated by linearly interpolating the coordinates from "snapshots" taken during the protein folding simulation, and so is not intended to be physically realistic.


This animated GIF image shows a synthetic folding trajectory for another small protein (a protein of unknown function from T. thermophilus) which in this case incorporates a small beta-sheet. This protein was predicted using FRAGFOLD3 during the 6th CASP experiment and the prediction had an RMSD of only 2.4 A from the experimental structure! Unlike the original FRAGFOLD, FRAGFOLD3 is able to model all of the details of the protein structure, including all of the side chain atom positions. This work is described in the following reference:

Jones D.T., Bryson K., Coleman A., McGuffin L.J., Sadowski M.I., Sodhi J.S., Ward J.J. (2005) Prediction of novel and analogous folds using fragment assembly and fold recognition. Proteins. 61 Suppl 7:143-51.

In this case, each frame of the animation was again generated by linearly interpolating the coordinates between each snapshot, but with additional structural refinement at each step to preserve the secondary structure features. Again, we do not claim this animation to be a physically accurate representation of the folding pathway for the protein. Nevertheless, by looking at such simulations we can make some educated guesses as to how proteins might fold in nature.

Note that the animations will run at full speed as soon as the whole file has loaded into your Web browser's cache.