Optimising the spatial structure of BLN protein models by means of "partial distortion''-quench cycles

The prediction of the spatial structure of a protein based on its amino acid sequence is a challenging problem.
Corresponding theoretical studies of the protein folding require highly efficient structure optimisation tools.
Here we investigate whether and to what extent the thermal cycling (TC) algorithm [1] is appropriate for determining low energy structures of the BLN protein model by J.D. Honeycutt and D. Thirumalai [2].
In our simulations for 46-, 58-, and 69-bead sequences, the TC algorithm reliably finds the global minimum within reasonable computing time.
In comparison to the multi-start local search and simulated annealing approaches, TC turns out to be far more efficient.

In the present work, the BLN model with rigid bonds is studied in detail for the first time.
Comparing these results to data for the extended model by Berry et al. [3], where stiff springs are substituted for the rigid bonds, we observe several level crossings when varying the spring constant, even for quite hard springs.

[1] A. Möbius et al. , Phys. Rev. Lett. 79 (1997) 4297.
[2] J.D. Honeycutt and D. Thirumalai, Biopolymers 32 (1992) 695.
[3] R.S. Berry et al. , Proc. Natl. Acad. Sci. USA 94 (1997) 9520.