On Protein Structure Prediction using the Hydrophobic-Polar Model on a Tetrahedral Lattice

Ryan Paul Gozum, Kristoffer Marion Mendoza

Adviser: Adrian Roy Valdez, PhD

Co-Adviser: Jaymar Soriano, MSc

Co-Adviser: John Justine Villar, MSc

Abstract

Predicting a protein’s final structure has been a long running problem in the field of bioinformatics. It is important to know how certain amino acid sequences fold to attain their minimal energy conformation, but to obtain these data is computationally expensive with today’s technology and physical limitations in computing. Modeling protein structures in free space allows for almost infinite possible conformations, and consequently, demands for similarly infinite resources. Different lattice models have been used to visualize close estimates of protein structures, but this entails tradeoffs of time complexity and accuracy, depending on each combination of model and energy minimization algorithm used.

In this paper, amino acid sequences are represented using the Hydrophobic-Polar (HP) model. The energy minimization is done using a proposed branch-and-bound method, and the resulting energy values are compared against existing benchmark sequences and with three lattice models considered, namely the cubic, triangular prism and the tetrahedral lattices.

This study aims to investigate the properties of certain lattices and understand their effects on the accuracy of finding solutions for the protein folding problem.