Prediction Of Missing Data For Anticancer Drug Sensitivity Based On The Matrix Complete

Some data missing in data bring a lot of inconvenience to analyze big data.As an important means of dealing with missing data,matrix completion has become an important research topic for big data analysis..Nowadays,cancer has seriously threatened human health and is one of the important culprits of death.Because of the heterogeneity of tumors,clinical trials are carried out before patients take drugs to select suitable drugs for cancer patients,in which the sensitivity test of cancer cell lines to antineoplastic drugs is a routine method.However,since drug sensitivity tests are often interfered and limited by experimental conditions,experimental equipment,experimental materials,etc.,the acquired drug sensitivity data is often missing.In the thesis,the matrix completion model has been researched.And the model has been applied to analyse anticancer drug response data for cancer patients.Firstly,a computational model was constructed to mine the similarity of cancer cell lines and antineoplastic drugs in anticancer drug response data based on the characteristics of biological data.And then,these information were integrated into the existing matrix completion model.A new low-rank matrix completion model was proposed to predicted the missing data of drug sensitivity.In addition,the relevant mathematical proof and the specific algorithm of model implementation are given.The matrix completion model was applied to two data from antineoplastic drug susceptibility database.Finally,the model is evaluated by ten-fold cross-validation and root mean square error.Compared with the results between the model in this thesis and existing models,the results showed that the model can effectively improve the prediction effect.