Tao Charles Y, Hoyt Jonathan, Feng Yan
Genome and Proteome Sciences Novartis Institutes for Biomedical Research 250 Massachusetts Avenue Cambridge, MA 02139, USA.
J Biomol Screen. 2007 Jun;12(4):490-6. doi: 10.1177/1087057107300707. Epub 2007 Apr 13.
High-content screening studies of mitotic checkpoints are important for identifying cancer targets and developing novel cancer-specific therapies. A crucial step in such a study is to determine the stage of cell cycle. Due to the overwhelming number of cells assayed in a high-content screening experiment and the multiple factors that need to be taken into consideration for accurate determination of mitotic subphases, an automated classifier is necessary. In this article, the authors describe in detail a support vector machine (SVM) classifier that they have implemented to recognize various mitotic subphases. In contrast to previous studies to recognize subcellular patterns, they used only low-resolution cell images and a few parameters that can be calculated inexpensively with off-the-shelf image-processing software. The performance of the SVM was evaluated with a cross-validation method and was shown to be comparable to that of a human expert.
有丝分裂检查点的高内涵筛选研究对于识别癌症靶点和开发新型癌症特异性疗法至关重要。此类研究中的关键一步是确定细胞周期阶段。由于在高内涵筛选实验中检测的细胞数量众多，且准确确定有丝分裂亚阶段需要考虑多个因素，因此需要一个自动分类器。在本文中，作者详细描述了他们为识别各种有丝分裂亚阶段而实现的支持向量机（SVM）分类器。与以往识别亚细胞模式的研究不同，他们仅使用低分辨率细胞图像和一些可以使用现成图像处理软件廉价计算的参数。通过交叉验证方法对支持向量机的性能进行了评估，结果表明其性能与人类专家相当。
