Recursive Cluster Elimination (RCE) for classification and feature selection from gene expression data

BMC Bioinformatics

Table 1 Summary results for the SVM-RCE, SVM-RFE and PDA-RFE method. Summary results for the SVM-RCE, SVM-RFE and PDA-RFE method applied on 6 public datasets. #c field is the number of clusters for the SVM-RCE method. The #g field is the number of genes in the associated #c clusters for SVM-RCE, while for the SVM-RFE and PDA-RFE indicates the number of genes used.

	Leukemia(I)			CTCL(I)			CTCL(II)			Head & Neck vs. Lung tumors (I)			Head & Neck vs. Lung tumors (II)			Prostate
	#c	#g	ACC	#c	#g	ACC	#c	#g	ACC	#c	#g	ACC	#c	#g	ACC	#c	#g	ACC
SVM-	2	12	99%	2	8	100%	2	8	91%	2	8	100%	2	9	100%	2	8	87%
RCE	3	32	98%	9	32	100%	9	34	96%	8	32	100%	6	32	100%	11	36	95%
	28	100	97%	32	101	100%	28	104	96%	28	103	100%	25	103	100%	32	100	93%
SVM-RFE		11	96%		9	89%		8	84%		8	92%		8	98%		8	93%
		32	96%		32	94%		32	85%		32	90%		32	98%		36	95%
		102	97%		102	100%		102	87%		102	90%		102	98%		102	94%
PDA-RFE		8	96%		8	92%		8	83%		8	89%		8	70%		8	94%
		32	96%		32	92%		33	81%		31	96%		32	98%		32	94%
		104	96%		104	95%		108	79%		109	96%		102	98%		104	90%

ISSN: 1471-2105