Discrimination techniques applied to the NCIin vitro anti-tumour drug screen: Predicting biochemical mechanism of action

摘要: The National Cancer Institute currently tests approximately 400 compounds per week against a panel of human tumour cell lines in order to identify potential anti-cancer drugs. We describe several approaches, based on these vitro data, the problem identifying primary biochemical mechanism action compound. Using linear and non-parametric discriminant procedures cross-validation, we find that accurate identification is achieved for 90 cent diverse collection 141 known compounds, representing six different mechanistic categories. demonstrate two-dimensional graphical displays terms initial three principal components (of original data) result suggestive visual clustering according action. Finally, compare classification accuracy statistical discrimination with obtained from neural network approach and, our example, results various approaches are similar.