Jeffrey Divergence Applied to Docking Virtual

摘要

Data analysis with high dimensionality and few samples implies a set of problems related with the Curse of dimensionality phenomenon. Molecular Docking faces these kind problems to compare molecules by similarity. LBVS-Ligand-Based Virtual Screening conducts studies of docking among molecules using their common attributes registered in specialized databases. These attributes are represented by high dimensionality boolean vectors where an bit set indicates the presence of an specific attribute in the molecule, whereas a zero bit, its absence. The discovering of new drugs through the comparison of these vectors involves exhaustive processes of matching among the vectors. In this work, it is proposed the use of Jeffrey divergence as a similarity measurement in order to find the best approximate virtual docking between distinct molecules, to reduce the computation time, and offset some of Curse of dimensionality effects. The results suggest the application of Jeffrey divergence on discovering of candidates to drugs allow to identify the best approximate matching among them.