Uma busca probabilística para o problema de geometria de distâncias moleculares

Abstract

The prediction of three-dimensional protein structures has been an area of intense research, addressed by several disciplines represented by bioinformatics. The present work proposes an innovative method for predicting protein structures based on a binary tree, called Frequency-Based Search (FBS). We carried out a statistical comparison of the efficiency of this method in relation to the Depth Search (DFS) method, using protein structures obtained by Nuclear Magnetic Resonance (NMR) available in the Protein Data Bank (PDB). The main objective is to evaluate the efficiency of these methods on subsequences of protein backbone atoms of sizes 5, 10, 15, 20 and 25, while investigating whether nature exhibits geometric preferences when folding. The computational results indicate that the FBS method outperforms the DFS method in at least 70% of the atom subsequences analyzed and suggest the existence of geometric preferences in proteins, as evidenced by the selected sample.