Publication Date

Spring 2011

Degree Type

Master's Project

Degree Name

Master of Science (MS)


Computer Science

First Advisor

Sami Khuri

Second Advisor

Mark Stamp

Third Advisor

Sarah Green


protein structure prediction hidden markov model


Genes that encode novel proteins are constantly being discovered and added to databases, but the speed with which their structures are being determined is not keeping up with this rate of discovery. Currently, homology and threading methods perform the best for protein structure prediction, but they are not appropriate to use for all proteins. Still, the best way to determine a protein's structure is through biological experimentation. This research looks into possible methods and relations that pertain to ab initio protein structure prediction. The study includes the use of positional and transitional probabilities of amino acids obtained from a non-redundant set of proteins created by Jpred for training computational methods. The methods this study focuses on are Hidden Markov Models and incorporating neighboring amino acids in the primary structure of proteins with the above-mentioned probabilities. The methods are presented to predict the secondary structure of amino acids without relying on the existence of a homolog. The main goal of this research is to be able to obtain information from an amino acid sequence that could be used for all future predictions of protein structures. Further, analysis of the performance of the methods is presented for explanation of how they could be incorporated in current and future work.