Integration of Support Vector Machine and Genetic Algorithm to Predict Disulfide Bonding Connectivity

• Main Authors: Shen-fu lin, 林聖富
• Other Authors: Yu-Ching Chen.
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/96435386473786672290

碩士 === 亞洲大學 === 生物與醫學資訊學系碩士班 === 100 === Disulfide bond is formed by two SH group of cysteines, and these two SH are oxidized to from a covalent bond. Furthermore, disulfide bond not only can be formed by sequence adjacent cysteine, but also is spatial proximity of cysteine, However, disulfide bond can help protein folding, stabilizing the protein structure, and have a great relationship with the regulation of protein function. In general, more disulfide bonds in a protein can make protein structure more stable, and not easily be destroyed. In this study, I combine the genetic algorithm (GA) and Support Vector Machine to predict disulfide connectivity. Genetic algorithm is used to select the feature vectors, and the feature vectors I used here includ cysteine-cysteine coupling, cysteine spacing patterns, position specific substitution matrix, AAindex, and, amino acid contents. The proteins I used to predict their disulfide connectivity are with two to five disulfide bonds, and the sequence similarity within proteins are equal and less than 30%. Finally, the accuracy of this method is between 62% to 67%.