基于药物晶体的立体形态参数辨识硫酸氢氯吡格雷的不同晶型

Identification of the polymorphs of clopidogrel bisulfate based on the steric morphology parameters of crystals

晶体的三维形态与晶型有着天然的相关性,本文基于晶体立体形态参数,建立以硫酸氢氯吡格雷的晶型辨识方法与模型。采用同步辐射光源X射线显微CT技术(synchrotron radiation X-ray microscopic CT technology,SR-μCT)和多层感知器(multilayer perceptron,MLP)神经网络数学建模相结合的方法,以两种晶型硫酸氢氯吡格雷晶体(clopidogrel bisulfate)与微晶纤维素丸芯的混合物为样本,通过对样本进行CT扫描并重构,构建三维结构模型,得到7组三维形态参数,再基于多层感知器神经网络算法建立数学模型,用于晶型的辨识和预测。所建模型对硫酸氢氯吡格雷晶型的预测成功率为92.7%,ROC曲线下面积为96.2%。通过描述晶体三维形态可以对药物晶型进行有效辨识。晶体的体积(volume)、顶角数(number of vertices)和表面积(area)对于硫酸氢氯吡格雷晶型的确定起决定性作用。

The crystal form of solid substance had intrinsic correlation with its three dimensional crystal morphology. Based on the characterization of the three dimensional crystal morphology of clopidogrel bisulfate, this research is to establish a model based on the three dimensional morphological parameters. The granular samples composed of polymorphs of clopidogrel bisulfate and microcrystalline cellulose （MCC） were scanned by synchrotron radiation X-ray microscopic CT technology （SR-μCT） and the three dimensional structural models for which were constructed. Seven groups of three dimensional morphological parameters were calculated. Finally, the mathematical model was established with the multi-layer perception （MLP） artificial neutral network methods to identify and predict the polymorphs of clopidogrel bisulfate. The success rate of the model prediction for the polymorphs of clopidogrel bisulfate was 92.7% and the area under the ROC curve was 96.2%.. The polymorphs of drugs could be identified and predicted through the numerical description of the three dimensional morphology. The volume, number of the vertices and the surface area were the major determinants for the identification of the polymorphs of clopidogrel bisulfate.