Predicting the Functions of Proteins from their Co-occurrences with Implicit and Explicit Functional Terms in Texts

Recent computational methods take advantage of the exponential explosion of biomedical literatures to predict protein functions. They do so by extracting information from the literatures that directly (i.e., explicitly)describe the functions of already annotated proteins. We observe that some biological terms pertaining protein functions may co-occur implicitly with proteins in biomedical texts. This has led us to believe that the methods that rely only on explicitly mentioned biomedical terms in texts may miss vital information about protein functions that is implicitly mentioned in the texts. Towards this, we propose in this paper an Information Extraction system called IPFI that employs techniques for predicting the functions of proteins from their co-occurrences in biomedical texts with both implicitly and explicitly mentioned biological terms pertaining functional categories. That is, IPFI uses a combination of explicit term extraction methods and logic-based implicit term extraction methods. It extracts explicit terms using Natural Language Processing techniques. It infers implicit terms by employing the inference rules of predicate logic. It triggers protein specification rules recursively. These rules are represented in the form of predicate logic's premises. We evaluated IPFI by comparing it experimentally with four existing methods. Results revealed marked improvement.