Lattice models with asymmetric propensity matrices for locationally informed protein structure prediction

Abstract

In this work we developed lattice models with novel scoring matrices used in conjunction with genetic algorithm for protein structure prediction. Specifically, we incorporated into the propensity scoring matrices the knowledge that hydrophobic residues should be near the core of the folded protein while hydrophilic residues should be on the outside, thus scored a conformation generated by the genetic algorithm by taking into account residues' positions in relation to the center of the conformation, which we define as the centroid of all residues. Results from the new scoring matrices show noticeable improvements, many significantly, over the standard HP lattice model and the more recent HPNX and hHPNX models.