论文部分内容阅读
目的在于定量预测雄激素受体干扰物活性,并确定最佳建模方法。选择150个分子作为数据集,随机选38个分子作为检验集,其它分子为训练集。每个化合物分子计算了193个分子参数。通过采用多元线性回归和主成分回归等方法,建立数学模型,并用验证集检验了所建模型的预测能力。结果发现逐步筛选法和主成分分析方法所建模型都表现出较强的预测能力(应用于检验集的相关系数分别为R=0.61,R=0.52)。以上研究将有助于新药雄激素受体抑制剂的筛选和开发。
The goal is to quantitatively predict androgen receptor antagonist activity and to determine the best modeling method. Select 150 molecules as the data set, randomly selected 38 molecules as the test set, other molecules for the training set. 193 molecular parameters were calculated for each compound molecule. Through the methods of multiple linear regression and principal component regression, the mathematical model is established and the predictive ability of the model built is verified by the verification set. The results showed that both the stepwise screening method and the principal component analysis method showed strong predictive ability (the correlation coefficients applied to the test set were R = 0.61 and R = 0.52, respectively). The above research will help the screening and development of new androgen androgen receptor inhibitors.