Functional Prediction And Similarity Calculation Of CircRNA Based On Multi-source Data

Circ RNAs are transcripts generated by reverse splicing,which are extremely stable and highly expressed molecules that also possess inter-species evolutionary conservation.They can regulate cellular processes and thus affect biological functions through different mechanisms of action,including acting as micro RNA sponges,interacting with RNA-binding proteins(RBPs),encoding specific peptides,and other ways.In recent decades,traditional biological experiments for circRNA function identification are highly accurate,but there are problems such as high experimental cost,practical expertise and low efficiency,which cannot be applied to large-scale circRNA function identification,so the functions of most circRNAs remain unknown.Therefore,using bioinformatics to explore circRNA function and functional similarity is an important part of circRNA research today.This thesis uses bioinformatics methods to predict circRNA function and calculate functional similarity based on multi-source data.The main research of this thesis is as follows.1.To address the problems such as high cost of circRNA function identification by traditional biological experiments,this thesis proposes a heterogeneous information network-based function prediction method,Bhin2 GO.circRNA interacts with micro RNA,protein and other molecules to influence biological functions,so this study incorporates circRNA related multiple data sources including circRNA,micro RNA and protein data to predict the gene ontology function of circRNAs.A global heterogeneous information network of circRNAs was first constructed.Then the topological information in the network was extracted using a feature extraction algorithm(Bhin2vec).Finally,a multi-label classification neural network was built for labeling prediction of circRNA function.In the independent test set,the F＿max reached 0.356.In addition,through ablation experiments,the results showed that the inclusion of micro RNA data significantly improved the prediction performance,consistent with the biological feature of circRNA as a micro RNA sponge.The comparison with traditional machine learning methods demonstrates the superiority of the classification model in this thesis.The final case study further illustrates that the present functional prediction method has good practicality and is useful as a guide for future circRNA functional identification screening studies.2.To address the problems of single source of circRNA similarity calculation data,this thesis proposes a multi-source data-based circRNA functional similarity calculation method-MSCFS.this method makes full use of the known association information between circRNA and disease to obtain circRNA disease functional similarity,the chaos game method to obtain circRNA sequence similarity from circRNA itself by The chaos game method obtains circRNA sequence similarity from the sequence information of circRNA itself and the onto2 vec algorithm obtains circRNA target gene similarity from the known association data between circRNAtarget gene-GO terms,and fuses the 3 kinds of similarities to obtain circRNA functional similarity.The performance of MSCFS was evaluated using circRNA-micro RNA association similarity and circRNA coexpression similarity,and the results showed a positive correlation with micro RNA association similarity(R=0.213)and circRNA co-expression similarity(R=0.8991),which can be used to infer the potential functions of circRNAs and can be used for circRNA-disease association prediction and discovery of circRNA similar functions.3.A platform for circRNA function prediction and functional similarity query was constructed.Based on the circRNA function prediction model and functional similarity calculation model constructed in this thesis,a circRNA research platform with a user-friendly interface and instructions for use is provided for researchers.Users can make prediction of circRNA function and circRNA functional similarity query through web side,and also download relevant data and model code for research.The construction of this platform provides a boost for circRNA researchers.