从蛋白质序列出发,我们首先将每条蛋白质序列分成等长的15段得到离散增量值、低频功率谱密度值、N端和C端的矩阵打分值和模体频数构成的组合向量表示相应的序列信息,用支持向量机算法,对六类酶家族类的亚类分别进行分类预测。预测的总精度依次为92.4%,92.9%,95.0%,97.3%,95.6%和97.9%。除转移酶和水解酶略低于前人,其余好于前人的预测结果。 Starting from the protein sequence, we first divide each protein sequence into 15 segments of equal length to obtain discrete increments, low frequency power spectral density values, matrix scoring values ​​of the N and C ends and a combination vector of the motif frequencies to represent the corresponding sequences Information, using support vector machine algorithm, the six categories of enzyme family subcategories were classified forecast. The total prediction accuracy was 92.4%, 92.9%, 95.0%, 97.3%, 95.6% and 97.9%, respectively. In addition to the transferase and hydrolase slightly lower than the previous, the rest better than the previous prediction.