Based On Biomedical Text Mining Technology Target Protein Prediction Of Natural Products

The prediction of small molecular lead compounds is an important topic in the drug research and development. This is because that the interaction between lead compounds and target proteins is a core issue of organic chemistry and drug research. This topic is always a hotspot due to the limitation of traditional biochemical experiment methods which are high cost and low throughput. With the rapid development of biomedical science, the number of relevant scientific literature is accumulated quickly. As a consequence, the main problem discussed in this thesis is how to make use of the literature information by information extraction (IE) and knowledge discovery (KD) for assisting the prediction of target protein.Information extraction (IE) is defined as extracting the information in the text into an organization form. The input of literature based IE system is the original text while the output is the structured information. The knowledge discovery (KD) is a process of recognizing the intelligible patterns which are effective, novel and useful.This thesis makes use of data mining algorithms for the prediction of target proteins and focuses on the following two aspects:Information Extraction:we present an overview of recent IE research in biomedical text mining and discuss the implementation of related techniques in detail. Meanwhile, we extract the co-occurrence information between target protein and drug from biomedical literature and develop an effective method for analyzing such information.Knowledge Discovery:we propose a modified framework which is based on KD for mining the potential relationships between the natural product and the target protein. In addition, we incorporate the literature information into support vector machine effectively, and verify its validity using some standard datasets.Overall, our work not only involves the discussion of machine learning algorithms but also an overview for related aspects in drug target prediction. Furthermore, the proposed models have been applied to a practical problem, that is, the efficient prediction of target proteins for the natural products.