Predictive Studies Of Disease-related MiRNA And Anticancer Drug Combination Based On Deep Learning

As an important non-coding RNA that can regulate gene expression,micro RNAs(miRNAs)play an important role in human life activities.The prediction of the potential disease-related miRNAs can help to understand the molecular pathogenesis of diseases and provide references for the diagnosis and treatment of diseases.In addition,cancer is a major disease facing humans.The single-drug treatment has problems such as drug resistance,toxicity and insignificant efficacy.The development of combination drugs can just solve the above problems to a certain extent.The method of using computational models for prediction can provide effective references for biological experiments and reduce the high cost caused by blind experiments.Aiming at the disease-related miRNAs prediction problem,this paper proposed a prediction model DBNMDA based on deep belief network.We constructed feature vectors to pre-train restricted Boltzmann machines for all miRNA-disease pairs and applied positive samples and the same number of selected negative samples to fine-tune Deep Belief Network to obtain the final predicted scores.The pre-training process of the model can learn the feature information of all samples,while the fine-tuning process can further improve the accuracy of the model,which is suitable for the problem of too few known samples in this study.In terms of performance evaluation,DBNMDA had achieved good prediction performance in global,local leave-one-out cross validation and 5-fold cross-validation.In the experimental part,three diseases were selected from three different perspectives for case studies,and most of the disease-related miRNAs predicted were verified by experiments.Aiming at the prediction problem of anti-cancer drug combination,this paper proposed a prediction model based on Siamese convolutional networks and random projection matrix.Firstly,the Siamese convolutional network and random projection matrix were used to process the features of the two drugs into drug combination features.Then the features of the cancer cell line were processed through the convolutional network.Finally,the processed features are integrated to predict the response data of the final drug combination to the cancer cell line through the full connection network.Compared with the traditional method of splicing drug features as drug combination features,the model in this paper can enhance the interpretability of drug combination features to some extent.In terms of performance evaluation,both the regression view of response data and the classification view of whether or not there is synergy,the model in this paper has achieved relatively excellent prediction performance.In addition,the introduction of copy number variation and mutation data of cancer cell lines can further enhance the prediction performance of the model.