Degree Type


Date of Award


Degree Name

Doctor of Philosophy


Genetics, Development and Cell Biology


Bioinformatics and Computational Biology

First Advisor

Drena Dobbs

Second Advisor

Robert Jernigan


RNA-protein interactions play important roles in fundamental cellular processes involved in human diseases, viral replication and defense against pathogens in plants, animals and microbes. However, the detailed recognition mechanisms underlying these interactions are poorly understood. To gain a better understanding of the molecular recognition code for RNA-protein interactions, this dissertation has three related goals: i) to develop methods for predicting RNA-protein interaction partners; ii) to develop an approach for predicting interfacial residues in both the RNA and protein components of RNA-protein complexes; and iii) to develop computational tools and resources for investigating RNA-protein interactions.

First, we present machine learning classifiers for predicting RNA-protein interaction partners. The classifiers use the amino acid composition of proteins and the ribonucleotide composition of RNAs as input to predict whether a given RNA-protein pair interacts. We show that protein and RNA sequences alone (i.e., in the absence of any structural information) contain enough signal to allow reliable prediction of interaction partners.

Second, we present RPISeq, a webserver that predicts the interaction probabilities of input RNA-protein pairs, using the above-mentioned machine learning classifiers. A comprehensive database of RNA-protein interactions, RPIntDB, is integrated with the webserver to allow users to search for homologous proteins and their known interacting RNA partners.

Finally, we perform an analysis of contiguous interfacial amino acids and ribonucleotides in RNA-protein complexes for which structures are known. We generate a dataset of bipartite RNA-protein motifs that can be used to predict interfacial residues in both the RNA and protein sequences of a given RNA-protein pair simultaneously. We show that taking binding partner information into account leads to higher precision in the prediction of RNA-binding residues in proteins.

Taken together, these studies have increased our understanding of how RNA and proteins interact.


Copyright Owner

Usha Muppirala



File Format


File Size

118 pages