Institute of Radioelectronics, Warsaw University of Technology, Nowowiejska 15/19, 00-665, Warsaw, Poland.
Adv Exp Med Biol. 2011;696:243-53. doi: 10.1007/978-1-4419-7046-6_24.
The problem of classifying chemical compounds is studied in this chapter. Compounds are represented as two-dimensional topological graphs of atoms (corresponding to nodes) and bonds (corresponding to edges). We use a method called contrast common pattern classifier (CCPC) to predict chemical-protein binding activity. This approach is strongly associated with the classical emerging patterns techniques known from decision tables. It uses two different types of structural patterns (subgraphs): contrast and common. Results show that the considered algorithm outperforms all other existing methods in terms of classification accuracy.
本章研究了化合物分类的问题。化合物用原子（对应于节点）和键（对应于边）的二维拓扑图表示。我们使用一种称为对比常见模式分类器（CCPC）的方法来预测化学-蛋白质结合活性。这种方法与决策表中已知的经典新兴模式技术密切相关。它使用两种不同类型的结构模式（子图）：对比和常见。结果表明，在所考虑的算法中，在分类精度方面优于所有其他现有方法。
