Detection of Gene Interactions Based on Syntactic Relations

[摘要] Interactions between proteins and genes are considered essential inthe description of biomolecular phenomena, and networks of interactionsare applied in a system's biology approach. Recently, many studies havesought to extract information from biomolecular text using natural languageprocessing technology. Previous studies have asserted that linguisticinformation is useful for improving the detection of gene interactions.In particular, syntactic relations among linguistic information are goodfor detecting gene interactions. However, previous systems give a reasonablygood precision but poor recall. To improve recall without sacrificingprecision, this paper proposes a three-phase method for detecting geneinteractions based on syntactic relations. In the first phase, we retrievesyntactic encapsulation categories for each candidate agent and target.In the second phase, we construct a verb list that indicates the nature ofthe interaction between pairs of genes. In the last phase, we determinedirection rules to detect which of two genes is the agent or target. Evenwithout biomolecular knowledge, our method performs reasonably well usinga small training dataset. While the first phase contributes to improverecall, the second and third phases contribute to improve precision. Inthe experimental results using ICML 05 Workshop on Learning Languagein Logic (LLL05) data, our proposed method gave an F-measure of 67.2% for the test data, significantly outperforming previous methods. We alsodescribe the contribution of each phase to the performance.

[学科分类] 基础医学