Discovery at a distance: Farther journeys in predication space

Abstract

In this paper we extend the Predication-based Semantic Indexing (PSI) approach to search efficiently across triple-predicate pathways in a database of predications extracted from the biomédical literature by the SemRep system. PSI circumvents the combinatorial explosion of possible pathways by converting the task of traversing individual predications into the task of measuring the similarity between composite concept vectors. Consequently, search time for single, double or triple predicate paths is identical once the relevant concept vectors have been constructed. This paper describes the application of PSI to infer double and triple predicate pathways connecting example pairs of therapeutically related drugs and diseases; and to use these inferred pathways to guide search for treatments for other diseases. In an evaluation of the utility of vector-based dual- and triple-predicate path search in a simulated discovery experiment, these approaches are found to be complementary, with best performance obtained through their application in combination.