Protein Function Analysis Using Amino Acid Residue Contacts and Support Vector Machines

ASIAIR > College of Computer Science > Department of Biomedical informatics > Theses & dissertations > Item 310904400/2537

Please use this identifier to cite or link to this item: http://asiair.asia.edu.tw/ir/handle/310904400/2537

Title:	Protein Function Analysis Using Amino Acid Residue Contacts and Support Vector Machines
Authors:	Lin Kuan Ming
Contributors:	Department of Bioinformatics
Keywords:	protein structure;SVM;proteomics;protein function
Date:	2004
Issue Date:	2009-11-06 14:31:31 (UTC+0)
Publisher:	Asia University
Abstract:	It is well known that various proteins carried out the functions of genes and these proteins can perform different functions due to their unique structures. Therefore, it is no doubt that there is a strong relationship between protein’s structure and function. The objective of this study was to analyze the frequencies of amino acid residue contacts to discriminate different types of protein functions. The structure data of 4 electron transport, 6 transcription regulation and 5 toxin related proteins were downloaded from a protein data bank (PDB). For each protein, the distances of amino acid pairs were calculated to determine the numbers and frequencies of the residue contacts. It was found that the frequencies of the residue contacts were not randomly arranged, in fact, the frequencies might represent the uniqueness of protein structure. Support vector machine (SVM) were used to analyze the frequencies data set to discriminate the three types of proteins. Only unique amino acid pairs out of total 210 types of pairs were selected to discriminate protein function. There were two ways to select the unique amino acid pairs: (1) to select amino acid pair which had the highest frequency from each protein (called “Top1 group”), (2) to select the top 10 frequencies amino acid pairs from the summation of total frequency of 15 proteins (called “Top 10 group”). SVM analysis showed that the frequencies of Top 1 and Top 10 groups could discriminate the 15 tested proteins into three functional groups. The results indicated that the proteins in the same functional groups had similar frequencies of residue contacts, therefore, the frequencies of unique residue contacts could be used to discriminate protein functions. Long range residue contacts were defined as the residue contacts that were far away each other in the sequence. For example, we used 1/4 and 1/2 long range residue contacts to represent the residue contacts that were away from each other in the sequence higher than 1/4 and 1/2 of total residues in that protein. SVM analysis of Top 1 group showed that the use of long range contacts was not helpful in discriminating protein functions.
Appears in Collections:	[Department of Biomedical informatics ] Theses & dissertations

Files in This Item:

File	Size	Format
	0Kb	Unknown	524	View/Open

Loading...