Leveraging artificial intelligence to predict ERG gene fusion status in prostate cancer

BMC Cancer

Table 1 Distribution of training and hold-out test datasets utilized for algorithm development

Dataset	Subset	Patients	ERG status		Gleason grade group
			Positive	Negative	1	2	3	4	5
TCGA cohort	Training set	235	123(52%)	112(48%)	41	68	60	32	34
Internal cohort	Training set	26	11(42%)	15(58%)	1	17	6	0	2
	Hold-out test set	131	60(46%)	71(54%)	0	67	31	4	29

Training subset includes initial training, cross-validation and testing sets. Hold-out test set refers to a separate subset of cases not included as part of the training subset. TCGA The Cancer Genome Atlas

ISSN: 1471-2407