基于协同过滤的药物重定位算法

Drug repositioning algorithm based on collaborative filtering

药物重定位是指发现已上市或批准药物的新用途,受到了广泛的关注.为此,提出一种基于协同过滤的药物重定位算法.首先,收集药物及疾病的描述信息以构建药物―疾病关联矩阵.其次,根据药物对疾病有适应症和有副作用的相关信息,设计了一种刻画药物之间趋同程度的度量方法,该方法同时考虑了不同药物在适应症和副作用上的相似度.然后,搜寻目标药物的近邻以预测药物对疾病的评分.最后,采用平均绝对偏差和覆盖率二项评价指标衡量系统的预测质量.另外,针对某种特定疾病,利用新的协同过滤模型预测药物在该疾病上的未评分项,根据预测的评分信息发现对该疾病有治疗作用的药物.实验结果表明,该算法不仅能提高系统的预测质量,而且能够发现有治疗作用的药物―疾病组合,验证了所提算法能有效地应用于药物重定位.

Drug repositioning,referring to find new uses for approved drugs or drugs already on the market,has attracted wide attention.In this paper,a novel algorithm of computational drug repositioning based on collaborative filtering is proposed.Firstly,the description information of drugs and diseases are collected to build the incidence matrix of drug-disease,which includes the related information of the indication and side effects of drugs on diseases.Then,according to the incidence matrix of drug-disease,a new measure method is designed to depict the convergence degree between drugs,and this method consides not only the indication similarity,but also the similarity of the side effecs.And then,according to the similarity among drugs,the Top-ksimilar drugs of the active drug are determined as its neighbors,and these neighbors are used for making the prediction on the target disease.Finally,the mean absolute error and coverage are selected as evaluation metrics to evaluate the prediction quality.Besides,airming at a given disease,the new model of drug repositioning based on collaborative filtering is utilized to predict unrated items of drugs.According to the predictive rating information,we discover potential drugs that have the indication on this disease.Experimental results show that the proposed algorithm can improve the prediction quality for the collaborativefiltering system,and can find some therapeutic combinations of drug-disease.Based on this,the conclusion is verified that the proposed algorithm can be applied to drug repositioning effectively.