Research And Application Of Drug Sensitivity Prediction Based On Molecular Functional Network

Today,with the prevalence of complex diseases,existing drugs are far from meeting the needs of humans to fight disease.Due to the continuous improvement and development costs of drug research,the development of innovative drugs has become a major challenge in the medical field.In recent years,with the continuous enrichment of databases of diseases and drugs,researchers have realized drug repurposing through correlation analysis of disease-related genes,drugs and drug target data,which is a new research and development idea in the field of pharmaceutical R&D.Innovative drug research and development costs,saving resources.It reduces the cost of research and development of innovative drugs and saves resources.Since most diseases are not single-gene defects,and often involve the destruction of coordination functions among the genes,this paper uses the biological function network to explore the relationship between drugs and diseases.With help of the HPRD,BioGRID,STRING and other databases,we constructed protein interaction network and used the random walk algorithm to predict drug sensitivity.The main work of this paper is as follows:1.A brief description of the current state of drug development and the background.We reviewed the current status of drug prediction research,the research status of drug relocation,drug disease related database,biological action network and random walk algorithm,which provides a theoretical basis for subsequent research.2.A drug sensitivity prediction method based on random walk algorithm was proposed.By screening disease-and disease-related genes and drug and drug target gene data,the sensitivity of the drug was predicted using a restart random walk algorithm.The results showed that the predictive model performed better in the lymphatic system of the blood and was relatively poor in autoimmunity,with predicted drug sensitivity 0.827 in the 10-fold cross-validation.Adding target information of a drug that can treat a disease to the disease gene information,or adding a gene of all diseases that can be treated by a certain drug to the target information of the drug,helps to improve the prediction of drug sensitivity.3.The analysis of the drug reuse and network functions.The restart random walk algorithm was used to obtain the relationship of each disease relative to each drug on PPI network,and the correlation coefficient was calculated.Through analysis of the 18564 group of diseases and drugs,there were 61 pairs of disease drugs with a correlation above 0.8,of which 53 disease drug pairs were known,and 8 pairs of drug disease relationships were not included in the data set.Through literature review,it was found that eight drugs have an effective on four diseases,and five of them have an effective effects on parkinsonian disorders.In this paper,we also found that the key link genes between parkinsonian disorders and therapeutic drugs are a-synuclein and protein tau.4.A drug sensitivity prediction method was proposed based on CREEDS database.The drug and disease treatment information were added separately for random walk,and the prediction accuracy was found to be 0.664 through ten-fold cross-validation.The results show that the data of disease and drug expression can detect the relationship between drugs and diseases to a certain extent,but the performance is not as good as the drug target and disease pathogenic gene data.The enrichment pathway of disease-related genes after random walk increased the immune-related pathways such as cell adhesion molecules and rheumatoid arthritis from the analysis of asthma-related genes.