|Title||Multistate protein design using CLEVER and CLASSY.|
|Publication Type||Journal Article|
|Year of Publication||2013|
|Authors||Negron C, Keating AE|
|Journal||Methods in enzymology|
|Keywords||Computational Biology, Computational Biology: methods, Protein Conformation, Proteins, Proteins: chemistry|
Structure-based protein design is a powerful technique with great potential. Challenges in two areas limit performance: structure scoring and sequence-structure searching. Many of the functions used to describe the relationship between protein sequence and energy are computationally expensive to evaluate, and the spaces that must be searched in protein design are enormous. Here, we describe computational tools that can be used in certain situations to provide enormous accelerations in protein design. Cluster expansion is a technique that maps a complex function of three-dimensional atomic coordinates to a simple function of sequence. This is done by expanding the sequence-energy relation as a linear function of sequence variables, which are fit using training examples. Generating a simpler function speeds up scoring dramatically, relative to all-atom methods, and facilitates the use of new types of search strategies. The application of cluster expansion in protein modeling is new but has shown utility for design problems that require simultaneous consideration of multiple states. In this chapter, we describe cases where cluster expansion can be useful, outline how to generate a cluster-expanded version of any existing scoring procedure using the software CLEVER, and describe how to apply a cluster-expanded potential to multistate protein design using the CLASSY method.