Molecular clustering via knowledge mining from biomedical scientific corpora

In this paper, an architecture is presented that allows the extraction of argumentation clauses that might exist in publications, in order to perform molecular clustering on referenced molecules. Grammar rules are defined and used to identify sentences corresponding to argumentation being present in publications. The references of those molecules are then compiled as lists that include their structure definition in SMILES format. These lists are given as input to virtual screening tools and then to a molecular clustering tool, with the ultimate goal to classify molecules that are known to be prone to specific diseases, thus leading to the discovery of new drugs.