Predicting Human Disease-associated CircRNAs Based On Kernel Ridge Regression Algorithm

Recently,a new type of RNA,circ RNA,has emerged in the field of RNA research,which is a non-coding RNA molecule with a special circular structure.The versatility,specificity and stability of circ RNA indicate that the molecule has great potential as a disease diagnostic marker.So,identifying the relationship between circ RNA and human disease will promote to understand the mechanism of the occurrence and development of complex diseases.In view of the limitations of traditional experimental methods,researchers are more willing to choose effective computational methods to identify new circ RNA-disease associations.In this paper,we constructed a novel computational model based on the kernel ridge regression to predict potential circ RNA-human disease associations.First,the integrated disease similarity is obtained by combining the semantic similarity and the Gaussian interaction profile kernel similarity of the disease.Next,the integrated circ RNA similarity is constructed by integrating the functional similarity,semantic similarity and Gaussian interaction profile kernel similarity of circ RNA.In order to reduce the complexity of the model,then the integrated circ RNA similarity and disease similarity were randomly selected and dimensionality reduction processing.Finally,the kernel ridge regression algorithm is used to construct two basic classifiers in the circ RNA space and the disease space,and use the average strategy to integrate them into one basic classifier.Since the random selection of features will obtain M basic classifiers,correspondingly M prediction scores will be obtained.This paper uses leave-one-out cross validation to evaluate the performance of this model,and it acquires AUC value of 0.938,which proves that this method is significantly better than state-of-the-art prediction method.In addition,the case studies of colorectal cancer further demonstrate the reliability of this method in discovering circ RNAs potentially related to human diseases,and also shows that this method has a better ability to predict the associations between unknown circ RNAs and diseases.