Content area
Full text
Abstract- Molecular recognition plays an important role in biological systems. In this paper, we are interested in receptor (protein) and ligand (non-protein) interactions what consists of noncovalent bonding such as aromatic stackings, hydrogen bonds, hydrophobic interactions and salt bridges. Understanding and predicting protein-ligand interactions is a complex task and in this paper we propose a model and algorithms to understand why different small molecules are recognized by a specific protein. Our model is based on graphs. Each protein-ligand complex is a bigraph where nodes are atoms and edges depicts interactions between protein and ligand. The proposed algorithms aim to detect conserved subgraphs in the dataset of graphs representing protein-ligand complex interactions. We also propose a visual interface where users can find general statistics about the dataset, the type of atoms and interactions established as well as select and analyze the generated patterns. We show an example of use of this methodology with Ricin and CDK datasets, both with their respective ligands.
Availability: A prototype of the visualization tool with the examples mentioned in the paper can be found at http://www.dcc.ufmg.br/~alexandrefassio/biocomp
Supplementary file: http://homepages.dcc.ufmg.br/~alexandrefassio/biocomp/files/suplementar.pdf
Keywords: Protein, Clustering, SVD, Graph, Pattern, Visualization
1. Introduction
Molecular recognition plays an important role in biological systems. It refers to interactions between two or more molecules through noncovalent bonding such as aromatic stacking, hydrogen bonding, hydrophobic forces and salt bridges. Solvent can play a dominant indirect role in driving molecular recognition in solution as well. The conditions responsible for the binding and interaction of two or more molecules are a combination of conformational and physicochemical complementarity [1]. Understanding and predicting protein-ligand interactions are essential steps towards ligand prediction, target identification, lead discovery and drug design [2].
In this paper we propose a model and algorithms to understand why different small molecules are recognized by a specific protein str u ct u re. It can be q u ite tricky beca u se of protein promiscuity [3], [4], [5] what leads to very dissimilar molecules being recognized by the same protein as its substrate or even acting as an inhibitor.
Despite the existence of several methods designed to predict protein ligands, few methodologies are devised to identify and describe what intelligible factors that imply in protein ligand affinity.
A straightforward...