|
1200 E. California Blvd.
Pasadena, CA
91125-9600
Mail Code: 114-96
Location: 131 Broad
(626) 395-6407
|
Research
Traditional
protein design methods select amino acid sequences that are consistent
with a single, fixed, main-chain structure by optimizing the identities
and conformations of the side chains. However, real proteins assume
an ensemble of conformational states, and modeling sequences on a
single backbone restricts our ability to select sequences that reflect
this essential characteristic. Similarly, it is difficult to generate
sequences with altered catalytic or binding specificity when only
one state is modeled at a time. These limitations of single-state
design have prompted interest in multi-state design (MSD) algorithms,
such as multi-state design Monte Carlo (MSD-MC). Previous results have
indicated that the Fast and Accurate Side-Chain Topology and Energy
Refinement (FASTER) algorithm is significantly more efficient than Monte
Carlo for single-state design. We developed and implemented a new optimization
algorithm for multi-state protein design based on FASTER, and compared
its performance to MSD-MC. Application of the MSD algorithms to several
difficult single-state design problems revealed that MSD-FASTER was
always able to find the optimal solution (as determined using well-tested
single-state design algorithms), whereas the best solutions found by
MSD-MC were sub-optimal. Both algorithms were able to successfully optimize
the straightforward core design of a 60-member NMR ensemble. Although
neither method was able to thoroughly sample a difficult surface design
of the same ensemble in a reasonable time frame, the solutions produced
by MSD-FASTER were significantly better. Finally, both algorithms were
applied to a negative design problem with three states, and both were
able to find an optimal sequence that stabilized one structure and destabilized
the other two, as desired. The performance of MSD-FASTER was nearly
100-fold better than MSD-MC in this case. We conclude that MSD-FASTER
is more efficient than MSD-MC for multi-state design in a variety of
contexts.
home | research | lab
members | publications | contact | links | group
only
© Calfornia Institute of Technology
|
