| Biomedicine contains a large amount of medical information and knowledge,but at the same time medical researchers are faced with the dilemma of "information overload".It has become an important means of innovating a new mode of health service and building a "Healthy China" to help the development of China’s medicine with artificial intelligence.In particular,the outbreak of "novel coronavirus" at the beginning of 2020 has put forward new requirements for promoting the application of clinical and scientific big data,strengthening the support of information technology in the field of biomedicine.This paper proposes a drug-disease treatment relationship discovery model based on named entity recognition and network link prediction,so as to further improve the utilization of biomedical literature and bring more hidden possibilities for disease treatment.In this paper,the data sources retrieved from the Sino Med Chinese biomedical literature database were obtained and preprocessed,and the "BIO" annotation method was used to manually annotate the data.The data was used to construct biomedical named entity recognition model based on depth study like BERT-Bi LSTM-CRF.The results show that the biomedical named entity recognition method based on the deep learning has a better performance.BERT-Bi LSTM-CRF model biomedical named entity recognition performs best,whose P value,R value,F1 value is 0.816,0.778,0.797 respectively.In this paper,a drug-gene bipartite network and gene-disease bipartite network were constructed and visualized.On this basis,the CN,Salton,Jaccard,HDI,HPI,LHN,AA,RA and PA indexes describing the partial information similarity of the bipartite network were extended and integrated.In this paper,the link prediction of drug-gene bipartite network and gene-disease bipartite networks was carried out according to the similarity index proposed above,and the prediction results were evaluated and analyzed.The results show that the network link prediction performance of RA index is better than other indexes of drug-gene bipartite network or gene-disease bipartite network.In this paper,the weights of drug-gene bipartite network’s edges and gene-disease bipartite network’s edges were predicted by means of the mixed and trimmed emotional dictionary.The results were used to construct drug-gene-disease associated network,which is used to get the disease treatment relationship of drug discovery.The results were evaluated and analyzed from three perspectives: drug discovery,disease discovery and drug-disease treatment relationship discovery for limited diseases.The drug-disease treatment relationship discovery method based on named entity recognition and network link prediction proposed in this paper can tap more potential possibilities for drug therapy of diseases from massive medical literature with the help of machine,and can find out the drugs that can treat some diseases and new applications of some traditional drugs in advance.On the one hand,it can avoid repeated experiments and save medical and social resources.On the other hand,it can control or slow down the occurrence and development of diseases in patients as soon as possible,and reduce the medical burden for patients.In particular,some research results in this paper can indicate the important role of some traditional Chinese medicine in the treatment of diseases,thus contributing to the development of China’s medical industry. |
