Phenotype prediction based on genome-wide DNA methylation data

BMC Bioinformatics

Table 7 Prediction performance (AUC) of MS-SPCA, using Normal data for training

	Normal HPV+ test1-5	Normal HPV+	Normal HPV-	CIN2+(a)	CIN2+(b)	Cancer
N.HPV+ 1	0.59/0.39		0.50/0.38	0.91/0.78	0.86/0.65	1/0.73
N.HPV+ 2	0.75/0.50		0.52/0.41	0.93/0.69	0.76/0.53	1/0.52
N.HPV+ 3	0.62/0.54		0.54/0.49	0.93/0.84	0.85/0.70	1/0.70
N.HPV+ 4	0.47/0.52		0.50/0.56	0.93/0.74	0.84/0.64	1/0.79
N.HPV+ 5	0.55/0.53		0.51/0.51	0.92/0.74	0.84/0.58	1/0.68
	Normal HPV- test1-5
N.HPV-1	0.63/0.31	0.62/0.53		0.91/0.69	0.71/0.55	1/0.80
N.HPV-2	0.14/0.47	0.63/0.59		0.91/0.67	0.71/0.57	1/0.76
N.HPV-3	0.42/0.53	0.61/0.60		0.92/0.67	0.67/0.56	1/0.72
N.HPV-4	0.36/0.41	0.63/0.59		0.92/0.67	0.67/0.56	1/0.76
N.HPV-5	0.39/0.55	0.50/0.47		0.82/0.51	0.50/0.39	1/0.67

Rows correspond to training datasets and columns to test datasets. The first number shows the performance of MS-SPCA, the second the performance of EVORA (mean value of 8 runs). Bold numbers show best predictions.

ISSN: 1471-2105