|
Title:
|
COMPARISON OF DISCRETIZATION METHODS OF FLEXIBLE-RECEPTOR DOCKING DATA FOR ANALYSES BY DECISION TREES |
|
Author(s):
|
Karina S. Machado, Ana T. Winck, Duncan D. Ruiz, Osmar Norberto de Souza |
|
ISBN:
|
978-972-8939-30-4 |
|
Editors:
|
Hans Weghorn, Pedro Isaías and Radu Vasiu |
|
Year:
|
2010 |
|
Edition:
|
Single |
|
Keywords:
|
Classification; Decision Trees; Discretization; Bioinformatics; Rational drug design; Molecular docking. |
|
Type:
|
Short Paper |
|
First Page:
|
225 |
|
Last Page:
|
229 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Molecular docking is an important step of the Rational Drug Design process where it is predicted the preferred conformation and orientation of a small molecule (ligand), to a biological macromolecule (receptor) in order to form a stable complex. To incorporate the receptor flexibility in docking we perform a series of docking simulations using in each one a receptor conformation constructed from a molecular dynamics simulation trajectory. This kind of approach generates vast amounts of data and is very computer-intensive. Aiming at reducing this computational demand and discovering information about the receptor-ligand interactions we apply the J48 implementation of C4.5 classification decision tree algorithm where the input mining files are based on the free energy of binding (FEB) that estimate the affinity of the receptor-ligand complex and distances between the receptor residues and the ligands. Since FEB is a continuous value and, in classification tasks, the target attribute must be categorical, we propose to compare three discretization methods for FEB: by equal frequency, by equal width and by mode and standard deviation and evaluate their impact in the generated decision trees. Moreover, using the induced decision trees we introduce a different approach to analyze receptor-ligand interactions from docking simulation results. |
|
|
|
|
|
|
