Study And Implementation Of The Classified Prediction System Of Compound Mutagenicity Based On SVM

With the advances of technology,more and more drugs are researched to against various diseases,the process of which has to cost plenty of material resources and manual power.Its R&D cycle is also too long.The whole process of drugs' R&D needs considering their five peculiarities,ADMET(absorption,distribution,metabolism,excretion and toxicity).In these five peculiarities,the toxicity of drugs includes mutagenic effects and it has close contacts with cancer.At the last stage of drugs' R&D,stage of animal and human trial will test types of drugs to human mutagenic effects.However,there are many experiments being given up on account of so many drugs have great damage to animal and human body in this stage.Thus,it has wasted the previous inputs of material resources and manual power.Recent years,pattern recognition in computer technology has developed rapidly,which is also applied in various field of society.The bioinformatics and drug R&D belongs to an important research direction of pattern recognition.The main function of this system is to use machine learning algorithm to make predictive classification on mutagenic effects of compounds,with disaggregated model analyzing the respectively related attributes of compounds.System offers a great deal of compounds and characteristic attribute as training set of classification model,which includes the results about studies on every institution researching on mutagenic effects of compounds.It also offers users the compound characteristic calculation,attribute selection,data cleaning,found of classification model,compound respective prediction,interpretation of result,destination file and so on.Researchers can analyze the key attributes of influencing compound mutagenic effects according to outcome of prediction.This system is used by Java language to exploit,by Spring MVC frame to make system architecture,by MySQL database to storage the data of compound characteristics and personal information,which has achieved data processing module,forecast classification module,result analysis module,system management module and personal information module.In the data processing module,the system calculates 1446 weft compound attribute descriptor of compounds and operates the characteristics of data on missing value processing and standardization,and then it uses the selection attribute algorithm of information gain,CFS and Relief to make dimension reduction process for attributes.In the forecast classification module,the system adopts SVM algorithm model and uses Adaboost arithmetic to make SVM iterative ascension so as so improve prediction accuracy of system.Through all kinds of cross validation and independent validation test set,the system can predict the mutagenic effects of compounds precisely,the precision reaching to 83.5%,which satisfies the needs of users in both function and property and reaches expected research results.