Research On Biomedical Texts Of Long Non-coding RNA

Long non-codingRNA(lncRNA)is a type of RNA that is more than 200 nucleotides in length and does not translate proteins.lncRNA has been increasingly revealed to play an important regulatory role in physiological and pathological processes in recent years.The abnormal expression of lncRNA is closely associated with many major diseases that seriously endanger human health(cardiovascular,Alzheimer and cancer,etc.).The study of lncRNA has become a new research hotspot in recent years.With the explosion of biomedical texts of lncRNA,it is of great application and challenge to automatically explore and mine the huge amount of lncRNA biomedical texts to further mine and extract more valuable information from them,and further provide more effective auxiliary tools for the subsequent study of lncRNA.Our paper uses text-based analysis and machine learning algorithms to analyze research hotspots and trends from multiple perspectives for lncRNA biomedical texts,providing researchers with an automatic and efficient model for lncRNA knowledge extraction,as well as a recommendation model for lncRNA paper submission and an online service platform.The main work of this thesis includes the following four aspects:(1)A comprehensive and systematic analysis of the available lncRNA biomedical texts.The publication trends and research hotspots of lncRNAs are reviewed,dynamic topic Model is used to mine lncRNA research topics as well as dynamic evolution,clustering algorithm is used to further refine lncRNA research,and the focus is on extracting existing lncRNA-disease associations,especially the association of lncRNA with cancer.(2)lncRNA-protein interaction(LPI)extraction model is constructed based on lncRNA biomedical text.LPI is essential for understanding the molecular mechanisms of lncRNA and inferring their functions.With the increasing number of biomedical texts,the extraction of LPI directly from texts will have great potential.Currently there is no tool for extracting LPI relationship from the literature.Automatic LPI mining from massive biomedical texts is a promising and challenging topic.Therefore,we construct a model for machine learning of identifying LPI based on multiple text features(semantic word vectors,syntactic structure vectors,position vectors and lexical vectors)and logistic regression classifier.(3)A journal recommendation algorithm model for lncRNA paper submission based on lncRNA biomedical texts is constructed based on a hierarchical attention network with multi-scale semantic feature fusion.At present,recommendation systems have been widely used in product recommendation,advertising recommendation and digital product recommendation,but rarely applied to journal recommendation,especially the paper recommendation system specifically for lncRNA.With the development of deep learning theory and its applications,deep learning-based paper recommendation systems are becoming more and more practical.To this end,we construct a journal recommendation model for lncRNA paper submission based on a hierarchical attention network with multi-scale semantic feature fusion.(4)A comprehensive online service platform for lncRNA biomedical text research is developed.Based on the first three main works in this thesis,we develop a comprehensive online service platform for lncRNA biomedical text research that consists of three main functional modules:(i)a display of lncRNA biomedical text analysis,(ii)an online service system based on multiple text features lncRNA-protein interactions extraction,(iii)A hierarchical attention-based multi-scale semantic feature fusion mechanism of lncRNA paper submission journal recommendation online service system.These three functional modules cover the hotspots of text-based lncRNA research and are largely convenient for researchers to use.Researchers can use this platform to get a visual overview of current lncRNA trends and research hotspots,easily extract LPI by submitting PMID,PMCID or biomedical texts in Pub Med,and get a candidate list of journals suitable for their research results by submitting the titles and abstracts of their papers.In addition,a manually screened and experimentally validated LPI corpus is available in our online service platform,which provides powerful data support for a comprehensive lncRNA-protein interaction extraction study.