基于矩阵填充算法的抗癌药物敏感性预测

Prediction of anticancer drug sensitivity based on matrix completion algorithm

在已有矩阵填充模型的基础上,融入癌细胞系的相似性信息与抗癌药物的相似性信息,提出了一种低秩矩阵填充模型,用于预测因受试验条件等影响而缺失的抗癌药物敏感性数据。将模型应用在CGP数据库中有缺失值的抗癌药物敏感性数据中,利用已有数据训练新模型的参数,并利用训练得到的最优参数对CGP数据库中的数据进行预测,得到最优近似值。结果表明,相较于其他模型,该模型能有效提高对缺失数据的填充效果。

The similarity information of cancer cell lines and the similarity information of anticancer drugs were incorporated into the existing matrix completion model.A low rank matrix completion model was proposed to predict the missing data of susceptibility to anticancer drugs due to the influence of experimental conditions and other factors.The model was applied to the data of anticancer drug sensitivity with missing values in CGP database.The parameters of the new model were trained by the existing data.The optimal parameters obtained by training were adopted to predict the data in the CGP database so as to gain the optimal approximate value.The results show that,compared to other models,the model can effectively improve the filling effect of missing data.