摘要
为了能区分不同的化合物,文中提出了一种基于多浓度-响应曲线的模式识别算法。在浓度-响应曲线上,可以提取到反映细胞毒性程度的特征值,然后使用固定了初始聚类中心的K均值算法来分类化合物。文中提出的特征提取、特征降维和聚类分析等方法能够实现对化合物的有效辨识。实验结果表明,所有测试化合物均能分类正确。
To determine distinct chemical properties,a pattern recognition algorithm using multiple concentration-response curves is developed in this paper. Features, which reflect the levels of cytotoxicity,are extracted from concentration-response curves. A K-means clustering method with deterministic initial centers is employed to classify the extracted features. The proposed method,including feature extraction,orthogonal projection and cluster analysis,can be readily automated,which enables relatively high throughput screening for chemical recognition. As a result,the tested chemicals are classified into several groups correctly.
出处
《信息技术》
2016年第2期165-169,共5页
Information Technology
关键词
基于时间的细胞响应曲线
特征提取
主成分分析
KKZ算法
K均值聚类算法
time-dependent cellular response profiles
feature extraction
principal component analysis
KKZ algorithm
K-means cluster