Gene–disease association extraction by text mining and network analysis

Abstract

Biomedical relations play an important role in biological processes. In this work, we combine information filtering, grammar parsing and network analysis for gene-disease association extraction. The proposed method first extracts sentences potentially containing information about gene-diseases interactions based on maximum entropy classifier with topic features. And then Probabilistic Context–Free Grammars is applied for gene-disease association extraction. The network of genes and the disease is constituted by the extracted interactions, network centrality metrics are used for calculating the importance of each gene. We used breast cancer as testing disease for system evaluation. The 31 top ranked genes and diseases by the weighted degree, betweenness, and closeness centralities have been checked relevance with breast cancer through NCBI database. The evaluation showed 83.9% accuracy for the testing genes and diseases, 74.2% accuracy for the testing genes.