Microarray-based cancer prediction using single genes

BMC Bioinformatics

Table 3 Comparison of single gene classifiers and standard classifiers

Parameter Dataset	Smallest p-value^a	t-test statistic^b	Fold change^c	# Significant gene^d	Accuracy (%) of standard classifiers^e	Accuracy (%) of single gene classifiers^f
Melanoma	1.37e-29	22.68	277.78	7263	97	96.5
Breast Cancer 1	8.10e-06	9.06	3.65	20	52.2	66
Brain Cancer	1.51e-04	4.06	21.73	15	67	78.5
Breast Cancer 2	3.10e-06	5.16	3.48	180	70	54
Gastric Tumor	7.34e-10	9.51	10	4798	92.2	84.5
Lung Cancer 1	2.51e-21	20.34	1923.48	7561	97.2	96.5
Lung Cancer 2	6.82e-35	24.72	505.16	3219	99	93
Lymphoma	1.50e-04	4.07	1.33	2	58.5	72.5
Myeloma	5.00e-07	5.23	4.49	172	76.5	67.5
Pancreatic Cancer	1.30e-06	5.37	5.88	58	61	79.5
Prostate Cancer	1.34e-21	12.53	12.82	812	89.3	89

Note:
^aThe minimum univariate t-test p-value for the genes significantly different between the classes.
^bThe absolute value of the t-test statistic corresponding to the left smallest p-value.
^cThe maximum fold change in the geometry mean of gene expression between the classes,
^dThe total number of genes significantly different between the classes at 0.001 significance level.
^eThe mean classification accuracy of the four standard classifiers.
^fThe mean classification accuracy of the two single gene classifiers.

ISSN: 1471-2105