Integrated analysis of independent gene expression microarray datasets improves the predictability of breast cancer outcome

Table 2 Cross-validation of two independent datasets

Profile	Validated by	Accuracy	AUC	Odds ratio
Rosetta	Stanford	74.2%	0.808	9.06
Stanford	Rosetta	70.7%	0.786	6.68
Overall		72.2%	0.795	6.71

SEP scores of validating patients were calculated with the 60-gene profile of the validated dataset. Patients were evenly classified into two groups using median score as threshold. Accuracy: percentage of correctly classified patients; AUC: area of SEP-based ROC curve; Odds ratio: the association between SEP and disease outcome based on Fisher's exact test.

ISSN: 1471-2164