Key LncRNA Prediction In Gene Expression Data Based On Machine Learning

Content:With the rapid development of genomics and bioinformatics,more and more studies showed that IncRNA is widely involved in biological regulation processes,especially plays an important role in the occurrence,development and prevention of human diseases.The study of feature mining and function prediction of IncRNA is becoming one hot spot in academia and industry field.In particular,predicting disease related key IncRNA which can reveal the cause and mechanism of disease,has very high biomedical value.However,IncRNA gene expression data has large size and complex structure,predicting key IncRNA is a difficult challenge.Machine learning provides us a feasible method of new ideas.Our topic focused on predicting key IncRNA in gene expression data,carried out related investigation,research,experiment and analysis,and used machine learning technology to design the improved BPSO-ELM key IncRNA prediction model.The main work of this paper is as follows:(1)Research the strategy of predicting key IncRNA in gene expression dataFrom the characteristics of IncRNA expression data and the research strategy of gene expression data,we analyzed and compared common genetic feature selection framework.And the research strategy of predicting key IncRNA in gene expression data is expounded step by step.(2)Build a efficient,accurate and adaptive key IncRNA prediction modelThis paper proposes a key IncRNA prediction model based on extreme learning machine and improved binary particle swarm optimization algorithm(Improved BPSO-ELM key IncRNA prediction model).LncRNA feature combination optimization problem is transformed into a binary particle swarm model.(3)Model development,validation and evaluationModel experiments were conducted on three typical datasets(Breast invasive carcinoma,Colon adenocarcinoma and Lung adenocarcinoma).The experimental results show that our prediction model can achieve more than 93.5%classification accuracy,and has better performance than other methods.