Prediction Of Drug Interactions Based On Complex Networks

Drug-drug interaction refers to the drug response that exceeds the expected effect after taking two or more drugs at the same time.It can be roughly divided into two categories: beneficial and harmful to the human body.The former can improve the efficacy of drugs and accelerate the recovery of patients,while the latter will reduce the efficacy and Even toxic side effects on the human body.In order to avoid the harmful DDI caused by the combination of drugs in clinical treatment,many scholars began to use calculation methods to predict the potential DDI between drugs.At present,the commonly used DDI prediction method is to determine whether DDI exists by calculating the drug similarity.However,this method is too complicated to calculate,the data is difficult to collect,and the universality is poor.In order to make up for the deficiencies of these methods,this paper proposes the following two network-based DDI prediction methods.The first method is to convert the DDI prediction problem into a link prediction problem,and use five kinds of node similarity methods to achieve the overall prediction of DDI.Its advantage is that it can obtain higher-precision prediction results without complicated calculations and algorithm design.In order to better test the universality of the method,this paper compares 5 methods in 6 different data sets.The results show that in the "L" and "N" data sets,the five link prediction methods all show extremely high prediction performance,and the prediction accuracy is between 0.8 and 0.9,while in the remaining 4 data sets,The prediction accuracy can also generally reach above 0.4.The second method is an extension of the first method.By dividing DDI into two types: synergy and antagonism,the DDI prediction problem is innovatively transformed into a classification problem,which realizes the classification and prediction of drug synergy and antagonism.This method combines the information of drug synergy and antagonism with the network topology for the first time,construct a series of node similarity features as classification features,and use support vector machine models to achieve drug synergy and antagonism prediction.The results show that the characteristics based on the network topology can effectively distinguish the synergy and antagonism.In the "L" network,the prediction accuracy of this method reaches 0.7,and the prediction accuracy between 0.2 and 0.5 can also be maintained in the coordination and antagonism relationship prediction of other networks.In summary,this article is a brand new attempt in DDI prediction work,and achieves higher performance DDI prediction using only the drug interaction network structure information.Compared with the traditional drug similarity method,the method in this paper is simpler to calculate and more convenient to implement,and has certain practical significance in the prevention of harmful DDI.