相互情報量とSVMを用いた酵素反応におけるEC番号の推定手法の開発  [in Japanese] Prediction of EC Numbers for Enzymatic Reactions Using Mutual Information and Support Vector Machine  [in Japanese]

Search this Article

Abstract

EC 番号は酵素反応を分類するための階層的なシステムであり,酵素活性の予測や反応の類似性の指標として用いられる.しかしながら,EC 番号は酵素活性から,手動で割り当てられるため,反応によっては不完全あるいは全く割り当てがなされていないものもある.そこで,本研究では MCS アルゴリズムと相互情報量を用いた EC 番号の推定方法を提案し,さらに機械学習ならびに相関係数で妥当な EC 番号を推定した.本手法はジャック・ナイフ法の結果において最大で感度 (Sensitivity) 83%,精度 (Precision) 84% を示した.また,複数の EC 番号が割り当てられている反応に対しても推定を行い,EC 番号の候補リストの妥当性を確認した.The computational prediction of protein catalytic functions is usually based on sequence similarities between enzymes. However, this method is questioned because small changes in key residues may alter the enzyme function. An alternative approach is to use classification systems such as Enzyme Commission (EC) numbers. Although EC numbers represent the hierarchical classification of an enzyme function and a catalyzed reaction, enzymatic reactions have incomplete or no EC numbers owing to the fact that they are manually assigned. In this study, we propose a new method to predict and assign valid EC numbers for unclear enzymatic reactions using a maximal common substructure (MCS) algorithm and mutual information (MI). In addition, we predicted EC numbers by using two methods, support vector machine (SVM) and correlation coefficient. Our proposed method yielded both high performance and high flexibility in predicting the EC numbers, performed sensitivity of 83% and precision of 84%, respectively. Furthermore, we predicted to the reaction assigned multiple EC numbers, and confirmed the validity of the candidate list of EC numbers.

Journal

  • 研究報告バイオ情報学(BIO)

    研究報告バイオ情報学(BIO) 2011-BIO-26(7), 1-6, 2011-09-06

Codes

  • NII Article ID (NAID)
    110008605711
  • NII NACSIS-CAT ID (NCID)
    AA12055912
  • Text Lang
    JPN
  • Article Type
    Technical Report
  • Data Source
    NII-ELS 
Page Top