A new framework for computational protein design through cost function network optimization

Abstract

The combined pipeline used to generate energetic models (based on a patched version of the open source solver Osprey 2.0), the conversion to CFN models (based on Perl scripts) and CFN solving (based on the open source solver toulbar2) are all available at http://genoweb.toulouse.inra.fr/~tschiex/CPD

Algorithms for protein design

“…For example, in a recent study [53••], Simoncini et al compared their provable optimization algorithm implemented in the Toulbar2 program [54, 55] vs. the heuristic simulated annealing (SA) algorithm implemented in the Rosetta program using a GMEC-model. They found that, in a set of 100 test protein designs, SA often fails to compute the optimal answer even after running hundreds of times.…”

“…Promising improvements in the performance of protein design optimization algorithms for the GMEC-model were recently reported by Thomas Schiex and co-workers [53••–55, 68]. Their protein design optimization algorithm exploits weighted constraint satisfaction (WCSP) techniques, including fast soft local consistencies for bounding, an advanced branch and bound implementation, and sophisticated ordering techniques, to compute the GMEC significantly faster than competing approaches [54].…”

“…Their protein design optimization algorithm exploits weighted constraint satisfaction (WCSP) techniques, including fast soft local consistencies for bounding, an advanced branch and bound implementation, and sophisticated ordering techniques, to compute the GMEC significantly faster than competing approaches [54]. These algorithms are implemented in the Toulbar2 program [53••–55], with support only for discrete flexibility models, and in the Osprey protein design program [39, 68] (with full support for continuous flexibility models, described in the next section [64•, 69••]).…”

“…Developing algorithms that improve over the best state-of-the-art algorithms will enable enhanced modeling and, in consequence, better designs. In addition, as shown in multiple studies [9••, 20, 21, 25, 47•, 56•] improving flexibility models in protein design can result in sequences that are lower energy, are more similar to those of native proteins, or perform better experimentally [9••]. Developing new algorithms that expand the equicomplexity curve for expanded flexibility models, especially continuous flexibility, is likely to result in better protein designs.…”

Computational Protein Design Using AND/OR Branch-and-Bound Search

“…Unfortunately, these heuristic methods can be trapped into local minima and may lead to poor quality of the final solution. On the other hand, several exact and provable search algorithms which guarantee to find the GMEC solution have been proposed, such as Dead-End Elimination (DEE) [6], A* search [21,22,7,35], tree decomposition [32], branch-and-bound (BnB) search [14,31,3], and BnB-based linear integer programming [1,18].…”