A Neural Network Method for Protein Classification

Author

Arie D. Jones, Indiana State University

Date of Award

Fall 12-1-2002

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Physics

First Advisor

Torsten Alvager

Second Advisor

Valentina French

Third Advisor

Gary Stuart

Abstract

The investigation detailed in this paper attempts to utilize a Leaming Vector Quantization network in order to classify a set of mitochondrial proteins based upon their amino acid sequence. The Learning Vector Quantization network uses a nearest neighbor approach to classification. Input vectors are fed into the network, which produces output vectors. Those output vectors are then matched by means of a distance bias to a corresponding classification vector. The learning and test sets consisted of thirteen similar mitochodrial proteins from seventy-two different species. This provided a pool of over nine hundred proteins to use. Half of the species were used to perform the learning and the other half was utilized for the testing phase. This network shows a high degree of accuracy while minimizing the computational cost of the system. I conclude that the network is an excellent network for the classification of proteins using solely their amino acid sequences.

Recommended Citation

Jones, Arie D., "A Neural Network Method for Protein Classification" (2002). All-Inclusive List of Electronic Theses and Dissertations. 3443.
https://scholars.indianastate.edu/etds/3443