Bioinformatics Advance Access originally published online on November 2, 2005
Bioinformatics 2006 22(1):122-123; doi:10.1093/bioinformatics/bti756
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ClaNC: point-and-click software for classifying microarrays to nearest centroids
Department of Biostatistics, University of Washington Seattle 98195, USA
 |
ABSTRACT |
|---|
 TOP ABSTRACT REFERENCES |
|---|
Summary: ClaNC (classification to nearest centroids) is a simple and an accurate method for classifying microarrays. This document introduces a point-and-click interface to the ClaNC methodology. The software is available as an R package.
Availability: ClaNC is freely available from http://students.washington.edu/adabney/clanc
Contact: adabney{at}u.washington.edu
Supplementary information: http://students.washington.edu/adabney/clanc/figure2.pdf
In Dabney (2005), I described the classification to nearest centroids (ClaNC) method of classifying microarrays. ClaNC is derived from a classical method called linear discriminant analysis (LDA). LDA-based classifiers are simple and have been shown to outperform many more complicated alternatives (Dudoit et al., 2002; Lee et al., 2005; Tibshirani et al., 2002). ClaNC customizes LDA to the microarray problem by incorporating a feature-selection step. Briefly, ClaNC uses regular t-statistics to rank genes by their ability to distinguish the classes, then applies a class-specific procedure for selecting the genes to include in the classifier.
The prediction analysis of microarrays (PAM) method of Tibshirani et al. (2002) is another example of an LDA-based classifier for microarrays. ClaNC error rates tend to be substantially less than their PAM counterparts, for a particular number of genes used in the classifier (Dabney, 2005). Furthermore, ClaNC error rates tend to be less variable than PAM error rates (Dabney, 2005).
ClaNC runs on top of R (R Development Core Team, 2005, http://www.R-project.org) and is available as an R package. Detailed installation instructions can be found at my website. Once invoked, the ClaNC interface will appear as in Figure 1. A typical analysis will proceed as follows. First, load data to be used in training the classifier; all data input takes place in the top frame, labeled Inputs. Second, specify the prior probabilities to be used and the range of active genes per class to consider in the middle frame, labeled Options. Choosing Equal priors will place prior weight 1/K on each class, where K is the number of classes. Choosing Class priors will place weight nk/n on class k, where nk is the number of training samples in class k, k = 1, 2, ..., K, and n is the total number of samples. The number of genes to use in each class is specified directly. By default, classifiers using 1, 2, ..., 10 genes in each class will be evaluated. To be clear, the classifier using 1 gene per class will choose K unique genes, one unique entry for each class. Third, train the classifier by clicking Train in the bottom frame. This will carry out 5-fold cross-validation to assign misclassification error rates to each classifier under consideration. Once completed, a plot will appear summarizing the results, as in Figure 2 (Supplementary material).
|
Fourth, choose a single number of active genes per class based on this plot. In the example of Figure 2, four genes per class (16 total unique genes) drop the error rate to zero for all four classes. To build the classifier based on four genes per class, select Single under Genes Per Class in the Options frame, and click Build. Fifth, if you would like to evaluate your classifier on independent test data, load the data with Test selected in the Inputs frame, then click Test. A summary of the results will be printed in the message box. Sixth, if you would like to predict the class of an unknown sample (or a collection of unknown samples), load the data with Predict selected in the Inputs frame, then click Predict. The predicted classes will be printed to the message box. Finally, you can Save your classifier to a text file. Table 1 shows the results of saving the classifier built above. See the manual for a more detailed discussion of the file format, but note that all components necessary to represent the classifier are present. This file can later be loaded into ClaNC by selecting ClaNC in the Inputs frame.
|
|
Acknowledgments |
|---|
This research was supported in part by the Cancer-Epidemiology and Biostatistics Training Grant 5T32CA009168-29 and NIH grant 1 U54 GM2119-03.
Conflict of Interest: none declared.
|
FOOTNOTES |
|---|
Associate Editor: John Quackenbush
Received on September 20, 2005; revised on January 10, 2005; accepted on October 27, 2005
|
REFERENCES |
|---|
|
TOPABSTRACTREFERENCES |
|---|
Dabney, A.R. (2005) Classification of microarrays to nearest centroids. Bioinformatics, 21, 4148–4154
Dudoit, S., et al. (2002) Comparison of discriminant methods for the classification of tumors using gene expression data. J. Am. Stat. Assoc, . 97, 77–87[CrossRef][Web of Science].
Lee, J.W., et al. (2005) An extensive comparison of recent classification tools applied to microarray data. Comput. Stat. Data Anal, . 48, 869–885[CrossRef].
R Development Core Team. R: A Language and Environment for Statistical Computing, (2005) , Austria ISBN 3-900051-07-0 R Foundation for Statistical Computing, Vienna.
Tibshirani, R., et al. (2002) Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc. Natl Acad. Sci. USA, 99, 6567–6572
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||