Construction And Research Of Deep Learning Platform For Gene Expression Prediction

With the development of machine learning and deep learning,more and more algorithms have begun to be applied to the field of biological information.In recent years,due to the vigorous development of microarray technology,human genetic information has been gradually discovered.After investigation and research,more than 20,000 genes have been discovered in the human body.Although the current genetic technology is constantly developing,it is still very expensive to carry out the whole genome of the human body.Studies by the National Institutes of Health have shown that gene expression in humans is usually highly correlated,and they have found that about 1,000 landmark genes in the human body contain information about the remaining 80% of the genes in the human body.In order to effectively utilize these iconic genes,approximately 1000 landmark genes can be used as input and modeled using machine learning or deep learning algorithms to predict remaining gene expression data.However,there are few platforms for gene expression prediction.The use of machine learning deep learning algorithms to model gene expression prediction and establish a gene prediction expression platform for researchers and learners is the main problem to be solved in this paper.The regression algorithms in machine learning mainly include linear regression,kernel ridge regression,support vector regression,etc.Deep learning mainly uses neural networks to do regression.Members of the LINCS project team used linear regression to predict gene expression.Yifei Chen et al.also use deep learning algorithms.However,its deep learning algorithm model was limited by the machine configuration at the time.Two neural networks were established for gene expression prediction.Although the accuracy is good,it takes a lot of time to train and predict.In this study,the deep learning model is mainly used,supplemented by machine learning models,such as linear regression,kernel ridge regression and support vector regression,and the gene expression prediction platform is built.The user only needs to input a small amount of gene expression value,ie The expression value of the marker gene is then predicted by the platform using the model that has been trained in the background,and the output of the 9520 gene expression values is fed back to the user.After preprocessing the genetic data set,use the scikit-learn learning package to establish linear regression,kernel ridge regression and support vector regression models,use multi-layer fully connected neural networks,and use Drop Out and other techniques to adjust parameters and establish deep learning.The model,after tuning the model parameters,saves the model to the local for platform loading.This paper elaborates on the model principle,modeling process and use used by the platform,and analyzes the different input of the user,such as localized Blast sequence alignment.Finally,the Django framework is used to build and develop the gene expression prediction platform,and the related data is persisted and saved to the My Sql database.This study aims to better help relevant researchers and learners to use it to assist gene expression research.