药物-疾病关系预测:一种推荐系统模型

发布时间：2019-02-11 16:07

【摘要】：目的药物重定位是指发掘已有药物新的治疗作用,然而具有潜在治疗作用的药物-疾病往往隐藏在数以百万计的关系对中。该研究基于医疗大数据分析,预测具有潜在治疗关系的药物-疾病关系对。方法将社交网络中推荐系统模型应用于药物重定位研究,并假设具有相似化学结构的药物可能具有相似的适应症。从开源数据库收集已知药物-疾病的治疗关系、副作用关系以及药物和疾病特征描述符,计算得到药物-药物的相似度和疾病-疾病相似度,再构建推荐模型将上述信息融合,并预测具有潜在治疗关系的药物-疾病,最终得到预测关系对的排序列表。结果列表排名前500的关系对中,有12.8%得到临床实验支持或综述报道,20%得到模式生物实验或细胞实验支持。结论相比于已有分类模型和随机抽样结果,本模型可明显提高具有潜在治疗作用药物-疾病的富集程度。
[Abstract]:Objective Drug relocalization refers to the discovery of new therapeutic effects of existing drugs. However, drug-disease with potential therapeutic effect is often hidden in millions of relation pairs. Based on the analysis of medical big data, this study predicts drug-disease relationships with potential therapeutic relationships. Methods the model of recommendation system in social network was applied to the study of drug relocalization, and it was assumed that drugs with similar chemical structure might have similar indications. From the open source database, we collect the treatment relation, side effect relation, drug and disease characteristic descriptor of known drug and disease, calculate the similarity between drug and disease, then construct the recommendation model to fuse the above information. The drug-disease with potential therapeutic relationship is predicted, and the ranking list of predictive relationship pairs is obtained. Results among the top 500 relationship pairs, 12. 8% were supported or reviewed in clinical trials, and 20% were supported by model biological or cell experiments. Conclusion compared with the existing classification model and random sampling results, this model can significantly improve the concentration of drug-disease with potential therapeutic effect.
【作者单位】：合肥工业大学计算机与信息学院;佛蒙特大学计算机系;同济大学生命科学与技术学院;
【基金】：国家自然科学基金资助项目(No 31100956,61173117) 国家高技术研究发展计划(863计划)资助项目(No2012AA020405)
【分类号】：TP391.3;R95

【参考文献】