Design and implementation of a cyberinfrastructure for RNA motif search, prediction and analysis

RNA secondary and tertiary structure motifs play important roles in cells. However, very few web servers are available for RNA motif search and prediction. In this dissertation, a cyberinfrastructure, named RNAcyber, capable of performing RNA motif search and prediction, is proposed, designed and implemented. The first component of RNAcyber is a web-based search engine, named RmotifDB. This web-based tool integrates an RNA secondary structure comparison algorithm with the secondary structure motifs stored in the Rfam database. With a user-friendly interface, RmotifDB provides the ability to search for ncRNA structure motifs in both structural and sequential ways. The second component of RNAcyber is an enhanced version of RmotifDB. This enhanced version combines data from multiple sources, incorporates a variety of well-established structure-based search methods, and is integrated with the Gene Ontology. To display RmotifDB’s search results, a software tool, called RSview, is developed. RSview is able to display the search results in a graphical manner. Finally, RNAcyber contains a web-based tool called Junction-Explorer, which employs a data mining method for predicting tertiary motifs in RNA junctions. Specifically, the tool is trained on solved RNA tertiary structures obtained from the Protein Data Bank, and is able to predict the configuration of coaxial helical stacks and families (topologies) in RNA junctions at the secondary structure level. Junction-Explorer employs several algorithms for motif prediction, including a random forest classification algorithm, a pseudoknot removal algorithm, and a feature ranking algorithm based on the gini impurity measure. A series of experiments including 10-fold cross- validation has been conducted to evaluate the performance of the Junction-Explorer tool. Experimental results demonstrate the effectiveness of the proposed algorithms and the superiority of the tool over existing methods. The RNAcyber infrastructure is fully operational, with all of its components accessible on the Internet