Stereo Signature Molecular Descriptor

We present an algorithm to compute molecular graph descriptors considering the stereochemistry of the molecular structure based on our previously introduced signature molecular descriptor. The algorithm can generate two types of descriptors, one which is compliant with the Cahn–Ingold–Prelog priority rules, including complex stereochemistry structures such as fullerenes, and a computationally efficient one based on our previous definition of a directed acyclic graph that is augmented to a chiral molecular graph. The performance of the algorithm in terms of speed as a canonicalizer as well as in modeling and predicting bioactivity is evaluated, showing an overall better performance than other molecular descriptors, which is particularly relevant in modeling stereoselective biochemical reactions. The complete source code of the stereo signature molecular descriptor is available for download under an open-source license at http://molsig.sourceforge.net.本研究基于我们此前提出的签名分子描述符（signature molecular descriptor），提出了一种可计算兼顾分子结构立体化学特征的分子图描述符的算法。该算法可生成两类描述符：其一符合Cahn–Ingold–Prelog（Cahn-Ingold-Prelog, CIP）优先规则，可处理富勒烯等复杂立体化学结构；其二为计算效率更优的一类，基于我们此前定义的有向无环图（directed acyclic graph），此类有向无环图可被扩展为手性分子图。本研究对该算法作为正则化器的运行速度，以及其在生物活性建模与预测任务中的性能进行了评估，结果显示其整体表现优于其他分子描述符，这一优势在立体选择性生化反应的建模任务中尤为关键。本立体签名分子描述符的完整源代码可在http://molsig.sourceforge.net以开源许可证协议获取下载。