Automated Protein Classification Using Rigidity Analysis



Journal Title

Journal ISSN

Volume Title



Proteins are one of the most important biological structures found in nature. Consequently, the ability to determine a protein’s function quickly and accurately is of considerable importance to the scientific community. Many modern computational techniques seek to solve this problem by using available structural information to classify proteins in order to infer protein function. Most current protein classification systems use a combination of automated techniques and manual curation. Even so, some levels for certain systems rely solely on expert analysis. Generally, in these circumstances, the characteristics of the group- ing are considered too broad or vague for automated techniques. The architecture level of CATH is one such example. This thesis explores the application of rigidity theory to the physical structure of proteins in automated protein classification by augmenting standard protein data (in the form of secondary structure information) with rigidity data. Machine learning algorithms are used to perform the classification. We evaluate the effect of the rigidity data on the accuracy of these algorithms. Specifically, we focus on three architectures within the Mainly Alpha class: orthogonal bundle, up-down bundle, and alpha/alpha barrel.



automated protein classification, rigidity theory, protein, bioinformatics, computational biology, machine learning