Differential Evolution for Protein Structure Prediction Using the HP Model - Foundations on Natural and Artificial Computation - page 322

Differential Evolution for Protein Structure
Prediction Using the HP Model

J. Santos and M. Dieguez

Computer Science Department, University of A Coruna, Spain
{ jose.santos,martin.dieguez } @udc.es

Abstract. We used Differential Evolution (DE) for the problem of pro-
tein structure prediction. We employed the HP model to represent the
folding conformations of a protein in a lattice. In this model the nature of
amino acids is reduced considering only two types: hydrophobic residues
(H) and polar residues (P), which is based on the recognition that hy-
drophobic interactions are a dominant force in protein folding. Given a
primary sequence of amino acids, the problem is to search for the folding
structure in the lattice that minimizes an energy potential. This energy
reflects the fact that the hydrophobic amino acids have a propensity to
form a hydrophobic core. The complexity of the problem has been shown
to be NP-hard, with minimal progress achieved in this category of ab ini-
tio folding. We combined DE with methods to transform illegal protein
conformations to feasible ones, showing the capabilities of the hybridized
DE with respect to previous works.

1 Brief Introduction to the Problem of Protein Structure

Prediction

Proteins only attain their sophisticated catalytic activities by folding into com-
plex 3D structures. The amino acid hydrophobicity plays an important role in
defining this folding process: for reasons of thermodynamics, most proteins fold
with hydrophobic side-chains pointing inwards to form a hydrophobic interior
and present a hydrophilic surface to the watery cytoplasm of the cell. Anfinsen
[1] showed that a protein in its natural environment folds into a unique three
dimensional structure, the native structure, independent of the starting confor-
mation. The thermodynamic hypothesis states that the native conformation of
the protein is the one with lowest Gibbs free energy. That native state of a pro-
tein plays an essential role in the functionality of a protein. As experimental
determination of the native conformation is still dicult and time consuming,
much work has been done to forecast the native conformation computationally.
There is an extensive research done on the direct prediction of the final protein
structure conformations (secondary and tertiary structures). In the case of sec-
ondary structure prediction (local regular elements such as helixes and strands),
machine learning methods such as neural nets and support vector machines, can
achieve up 80% overall accuracy in globular proteins [7].