User assisted substructure extraction in molecular data mining

Abstract

In molecular fragments mining, scientists use both manual techniques and pure computer based methods. In this paper, we propose a novel molecular fragment mining approach that incorporates interactive user assistance to speed up and increase the success rates in traditional fragment mining processes. The proposed approach visualizes 3D molecular data in 2D form that can be easily interpreted by a human expert who evaluates and filters the 2D molecular images manually. The proposed approach differs from others in literature as it does not search substructures including specific atoms like graph mining methods do. Instead, user assisted approach highlights significant substructures with specific properties and topologies graphically. Initial experiments indicate that by the use of user assisted approach, active and inactive fragments of compounds are quickly determined for drug design with high success rates.