α-淀粉酶Amy7C及其突变体催化常数的定量预测

Quantitative Predicting Kcat of α-amylase Amy7C and Its Mutants

关键词:  氨基酸属性  α-淀粉酶  催化常数  预测

Key words:  amino acid property  α-amylase  Kcat  prediction

摘要:

[目的]利用α-淀粉酶Amy7c及其突变体的氨基酸信息,预测该酶的催化常数(Kcat),并筛选出能预测α-淀粉酶Kcat最具效果的氨基酸属性。[方法]先以20-1前馈反向传播的神经网络为模型,完成535种氨基酸属性对α-淀粉酶Amy7C及其突变体催化常数的拟合。再将α-淀粉酶Amy7C及其54个突变体的数据分为2组,用35个酶作为训练组进行拟合,20个酶作为验证组进行检验。最后,对8种不同层次及神经元个数的模型进行比较。[结果]110个氨基酸属性可实现20-1神经网络模型收敛,表明这些氨基酸属性可用于预测α-淀粉酶的催化常数,不同指标的预测效果不同。多模型的分析结果显示,不同模型对训练组R值的结果具有显著性差异,而对训练组P值、验证组R值和验证组P值结果无显著性差异。[结论]氨基酸分布概率等属性可以用于预测α-淀粉酶催化常数。四层神经网络模型是预测α-淀粉酶催化常数的相对理想的模型。

Abstract:

[Objective] The Kcat of α-amylase Amy7C and its mutants was predicted using amino acid information,and the most suitable amino acid property for predicting Kcat of α-amylase was selected.[Methods] 20-1 feedforward backpropagation neural network was used to screen 535 amino acid properties as predictors to predict the Kcat of α-amylase Amy7C and its 54 mutants,which were divided into two group,35 of them served as training group for fitting,and the other 20 were treated as validation. Eight models for different layers and numbers of neurons were also compared.[Results] 110 amino acid properties,which converged during fitting in the 20-1 neural network model,could be used to predict the Kcat. Different amino acid properties presented different predicting effect. The multi-model results showed that there was significant difference between R values in training groups,but there was no significant difference between P values in training groups,as well as R and P values in validation groups.[Conclusion] Some amino acid properties such as distribution probability could be used to predict the Kcat of α-amylase,to which four-layer neural network reveals the relative ideal model.