CLEAVER : A publicly available web site for supervised analysis of microarray data

Soumya Raychaudhuri*, Patrick D. Sutphin**, Joshua M. Stuart*, and Russ B. Altman*†.

Departments of Medicine (Biomedical Informatics)* and Radiation Oncology**, Stanford University, Stanford, CA.† Correspondence should be addressed to R.B.A (e-mail: rba@stanford.edu).

Supplement: typical output from Cleaver Classification for a variety of different problems. Cleaver Classification algorithm is an algorithm developed for supervised learning on micro-array data. It will be made publicly available in the form of an internet web site along with other statistical procedures. The table below has references, as well as links to the files that were used for each test case, the graphical and text output for each run, and summaries of performance. If a case is given a positive score - it is predicted to be a member of the positive set, otherwise a member of the negative set; the magnitude of the score indicates the confidence with which the assignment is made. The individual features are also scored to indicate the relative importance of each feature in prediction. The graphical output contains only the most influential features. All of the runs below were the result of cross-validation on the listed data.

CLEAVER web site

Table 1

							Cross Validation Accuracy
Problem	Gold Standard Reference	Positive Set	N	Negative Set	N	Features	Penalty = 0.1	Penalty = 10
Acute Leukemia (Golub)	Golub	Lymphoid data	47	Myeloid data	25	7129 genes	95.83% gif output text output	97.22% gif output text output
Lymphoma (Alizadeh)	Alizadeh	Diffuse Large Cell Lymphoma data	42	Non-DLCL data	54	4026 genes	95.83% gif output text output	94.79% gif output text output
Ribosomal Genes (Eisen)	Mewes (MIPS catalogue)	Ribosomal Genes data	121	Other Genes data	2346	79 exp.	99.23% gif output text output	95.18% gif output text output
Sporulation Promoters (Chu, Spellman , De Risi)	Chu, Mitchell	early genes (urs1 promoters) data	13	middle genes (mse promoters) data	23	103 exp.	97.22% gif output text output	94.44% gif output text output
VHL expression/hypoxia profile (Scherf, Ross)		VHL induced data	17	VHL repressed data	15	6828	75% gif output text output	75% gif output text output

 For any queries or suggestions contact sxr@smi.stanford.edu or sutphin@smi.stanford.edu

Golub, T.R., et al. Science 286, 531-537 (1999).

Alizadeh, A.A. et al. Nature 403, 503-511 (2000).

Eisen, M.B., Spellman, P.T., Brown, P.O., Botstein, D. Proc. Natl. Acad. Sci. USA 95, 14863-14868 (1998).

Mewes, et al. Nucleic Acid Res. 25, 28-30 (1997).

Spellman P.T. et al. Mol. Biol. Cell 9, 3273-3297 (1998).

Chu, S. et al. Science 282, 699-705 (1998).

DeRisi J.L. et al. Science 278, 680-686 (1997).

Chu, S., Herskowitz, I. Mol. Cell 1, 685-696 (1998).

Mitchell, A. Microbiological Reviews 58, 56-70 (1994).

Schref, U, et al. Nat. Genet. 24, 236-244 (2000).

Ross, D.T., et al. Nat. Genet. 24, 227-235 (2000).