An automated framework for hypotheses generation using literature.

Area of Science:

• Biomedical research
• Computational biology
• Bioinformatics

Background:

• Scientists face challenges in generating hypotheses due to the overwhelming volume of biomedical literature.
• Junior investigators struggle to formulate and validate hypotheses against existing research.
• Existing literature mining tools have not fully addressed the need for efficient knowledge harvesting in disease research.

Purpose of the Study:

• To develop a hypothesis generation framework (HGF) for identifying semantic associations between entities in biomedical literature.
• To bridge the gap between invested research efforts and the effective harvesting of published knowledge.
• To assist researchers in formulating and corroborating hypotheses by uncovering direct and indirect associations.

Main Methods:

• The HGF was designed using scalable computational models of disease-disease interaction.
• Integration of mapping ontologies with latent semantic analysis to capture crisp associations.
• Development of a framework to make assertions about entities, such as disease-factor associations.

Main Results:

• Pilot studies demonstrated the HGF's ability to capture direct and indirect "crisp" associations.
• Comparative analysis with curated expert knowledge validated the framework's assertions.
• The HGF provides on-demand knowledge discovery for hypothesis generation.

Conclusions:

• The HGF is a fast, efficient, and robust tool for generating novel hypotheses about disease-associated factors.
• A web service application is under development for wider dissemination of the HGF.
• Large-scale validation studies are ongoing with domain experts to confirm computed associations.