Design And Implementation Of The Classification Prediction System Of Hepatotoxicity Based On Ensemble Learning Methods

During the process of drug development,people usually consider the following five factors: absorption,distribution,metabolism,excretion and toxicity,and toxicity mainly includes carcinogenicity,teratogenicity,mutagenicity and hepatotoxicity.The liver is a vital organ of the body,and plays an important role in metabolism as it is sensitive to harmful substance.Besides,Drug-induced liver injury(DILI)is one of the leading causes of drug failure in trials and withdrawal from the market.Thus,determining the hepatotoxicity of compounds is essential.Over the past decades,several traditional approaches have been developed to assess the risk of DILI,both in vivo and in vitro.However,these studies are complicated,time-consuming,expensive,and may not yield high correlations between experimental results and effects observed in humans.With the drawbacks of traditional methods become more and more significant and the rapid development of computer technology,more and more toxicologists begin to use computational approaches to predict hepatotoxicity,which mainly use machine learning algorithms to predict the hepatotoxicity of compounds from chemical structure properties based on several datasets.In silico approaches are fast,low-cost and easy to implement,which are recognized as alternative methods to assess the DILI.This system modeling process mainly relies on the R language.Based on a large number of compounds collected from various literatures and various drug organization,we used three machine learning algorithms and 12 molecular fingerprints to produce 36 base models.Then we chose the best ensemble model by fusing these base models via averaging.At last,we got the ensemble model which achieved an average accuracy of 71.1±2.6%,sensitivity of 79.9±3.6%,specificity of 60.3±4.8%,and area under the receiver operating characteristic curve(AUC)of 0.764±0.026 in five-fold crossvalidation and an accuracy of 84.3%,sensitivity of 86.9%,specificity of 75.4%,and AUC of 0.904 in an external validation.It shows a high level in the field of hepatotoxicity prediction.This system is built under the Linux operating system.The front end is built by Html+Css+JavaScript,while the background is built by Apache+PHP+MYSQL,and we use the best ensemble model to predict the hepatotoxicity of the input compounds.We have implemented the login and registration module,the data preparation module,the prediction module,the result analysis module and the system management module.These results are clearly displayed in the system,and we also provide users with results query,results download and user management.The system has achieved our expectations in the functions and performance.