| Early screening of drugs usually requires steps such as drug design,cell assays,animal experiments,and clinical trials.The development and screening of a new drug usually costs nearly ten years and hundreds of billion Yuan,and eventually these costs are often wasted due to the incompatibility,inefficiency,toxicity and other reasons.Therefore,the study of a new drug effect prediction based on the combination of gene expression data from human cells and deep learning can reduce the unnecessary time and money waste in early screening of drugs,and contribute to new drug exploration success.This thesis uses the shortest path finding algorithm to capture the context from gene interaction network,and employs the captured gene context for the training of the machine learning algorithm to generate gene vectors that represent the function and meaning of each gene.Based on these gene vectors and gene expression data of cells,the cell vectors,that can represent the state of many aspects for each cell,are finally integrated,and the drug effect on cells can be determined by these changes in the cell state before and after drug treatment.In this way,we can determine whether drugs are suitable for human cells from a wider perspective,like the state changes of entire cell.In order to predict the changes of gene expression in human cells,this thesis constructs a deep learning model with molecular structures of drugs as input and cell-state changes as output,and applies the trained model to predict the effect of cell state caused by a new small molecule drug that the model has not learned,and finally the model achieved a good predictive performance.It proved the feasibility of deep learning for drug early screening and gene expression states prediction.In this research,the integration of deep learning,gene interaction network data,gene expression data and small-molecular compounds structure data can help us to predict the effects of drugs on human cells,thus allowing us to screen more effective,adaptable and low-toxic drugs to human cells.It also provides new ideas for the virtual screening of drugs. |
PDF Full Text Request