Mining And Analysis Of Adverse Drug Reactions In FAERS Database

With the development of medical and health industry,the number of new drugs is increasing.However,the drugs bring the risk of adverse reactions while in the treatment of diseases.Because drugs through little cases,and the study time is too short before putting it into market.The range of the test people in age is too narrow and too tight control to drug targets,including test design is too simple,many potential drug reactions are difficult to be found.It does cause a serious threat of life and property of the patient.The purpose of this study is to investigate and analyze the relevant data of the US Food and Drug Administration’s Drug Event Reporting System(FAERS).The aim is to carry out safety monitoring of listed drugs and to find out the suspected reaction signals and provide reference to the relevant institutions,to make more drugs in the clinical to use accurately and reduce the incidence of adverse drug reactions.In this paper,we selected all the data of the FAERS database from the fourth quarter of 2012 to the first quarter of 2016,and the data were cleaned with the Observational Health Science and Informatics Research Center(OHDSI).At the same time,it also completed the data extraction in the process.This experimentbasically solutes the spontaneous reporting data system data problemsthrough a series of data processing.Currently,the methods to monitordrug safety are basically disproportionality analysis,these methods take the DRUG-ADE pair’s relative risk ratio(relative risk,RR)to analysis.In this paper,the Reporting Odds Ratio,Proportional Reporting Ratio,Bayesian Confidence Propagation Neural Network and the Multi-item Gamma Passion confidence neural are competed based on the FAERS database.Those four methods were analyzed the reports of suspected adverse reactions.In addition to those four classical drug safety monitoring algorithms.this paper presents a three-component mixture model based on Multi-itemGamma Passion confidence neural.It creates the original hypothesis as the background incidence of adverse drug reactions and uses the local false discovery rate(lFDR)to rank the adverse reaction signal.What’ s more,the maximum likelihood estimation and the particle swarm algorithm based on the minimum spanning tree topology are used in this paper to get the parametric solution in the part of model.Analysis shows that the ternary mixed model and the local false discovery rate is slightly superior to the traditional drug adverse reaction signal in ranking.While through the experimental study,this paper found that different methods of different reported frequency of adverse drug reaction detection sensitivity is different.It can be combined in the actual problem of each method.The subject has practical significance in nowadays...