PT Unknown
AU Jaume Gibert
   Ernest Valveny
   Horst Bunke
TI Graph of Words Embedding for Molecular Structure-Activity Relationship Analysis
BT 15th Iberoamerican Congress on Pattern Recognition
PY 2010
BP 30–37
VL 6419
DI 10.1007/978-3-642-16687-7_9
AB Structure-Activity relationship analysis aims at discovering chemical activity of molecular compounds based on their structure. In this article we make use of a particular graph representation of molecules and propose a new graph embedding procedure to solve the problem of structure-activity relationship analysis. The embedding is essentially an arrangement of a molecule in the form of a vector by considering frequencies of appearing atoms and frequencies of covalent bonds between them. Results on two benchmark databases show the effectiveness of the proposed technique in terms of recognition accuracy while avoiding high operational costs in the transformation.
ER