| Prediction of protein structure depends on the accuracy
and complexity of the models used. We represent polypeptide chains by a
sequence of rigid fragments that are concatenated without any degrees of
freedom. Fragments chosen from a library of representative fragments are
fitted to the native structure using a greedy build-up method. This gives
a one-dimensional representation of native protein three-dimensional structure
whose quality depends on the nature of the library. Each library is characterized
by the quality of fit (accuracy) and the number of allowed states per residue
(complexity). Our goal is to find representations that are both accurate
and economical (low complexity). We recently constructed a library that
outperforms earlier libraries in this regard: with 10 states per residue
we approximate native protein structure to 1 angstrom compared to over 20
states per residue needed previously. We plan to extend the above described
models to approximate the full backbone chain, rather than only the C-alpha
atoms approximated in the current model. This can be done by modeling each
of the library fragments by a series of dihedral angles. By exploiting the
correlation between angles of neighboring residues along the polypeptide
chain, one can produce simple and accurate libraries. The
library fragments can be used to model, and possibly predict, protein
loop structure. The library of fragments, along with the construction
scheme for appending fragments, allows the enumeration of all possible
structures of a fixed length. Unfortunately, the number of structures
is exponential in the length of the chain. Loops are fairly short, and
thus the total number of structures is manageable. To further improve
the performance we may construct a loop simultaneously from its two ends.
Decoy structures can also be modeled with these libraries,
by randomly sampling the structures space. As noted, there are exponentially
many possible structures deeming uniform random sampling ineffective.
Alternatively, we can bias the sampling of structures by the secondary
structure prediction of the protein considered. We believe that this method
can be utilized as an effective starting point for other decoy refining
schemes
|
