A Novel Approach for Adverse Events Detection

Adverse events detection is critical in many fields, e.g., adverse drug event (ADE) detection in medical field. ADE is an unexpected and harmful consequence of drug usage. Many researchers found that identifying the correlation between the use of drugs and adverse events from biomedical literature can contribute a lot to drug safety monitoring. In this paper, we propose a novel approach based on biomedical literature to detect ADE. We first construct a graph using candidate ADE extracted from biomedical literature, and then propose a method to select critical vertices from the graph as core vertices with a clustering algorithm to group these core vertices to build subgraphs. Lastly, the correlations between drugs and events are calculated based on the subgraphs for ADE detection. Experiments show that our approach is highly feasible.