Hypothesis Driven Secondary Network Mining from PubMed Using Epiphany

Incomplete knowledge about biological objects and a universal set of rules can be used to guide the exploration of PubMed literature to aggregate knowledge for secondary network inferencing. Epiphany is a novel text mining architecture in which users are able to express what is known about an object, what general rules apply for conclusion drawing, and what hypothesis is being expected. Once stated, Epiphany mines the PubMed abstracts first to generate a possible primary weighted and annotated interaction network, and then generate a secondary and higher level interaction network from a subset of more relevant publications. This candidate network serves as a basis for secondary network prediction as a final step. This paper describes the proposed architecture of Epiphany as a higher level text mining tool in life sciences and presents a case study in which the mechanism of Dengue virus has been predicted solely from literature mining.