DNA methylation profiles capturing breast cancer heterogeneity

BMC Genomics

Table 1 The confusion matrix of Random Forest classification using the proposed methylation signature

Signature	Data type	Dataset		TN	nonTN	F1 score		MCC
Signature	Data type	Dataset		TN	nonTN	F1_TN	F1_nonTN	MCC
pCpGs	Methylation	GSE72245	TN	7	2	0.67	0.90	0.58
			non-TN	5	32
pDMGs	Methylation	TCGA	TN	11	43	0.29	0.93	0.26
			non-TN	12	384
pDMGs	mRNA	TCGA	TN	44	25	0.70	0.97	0.67
			non-TN	12	519
PAM50	mRNA	TCGA	TN	60	9	0.84	0.98	0.82
			non-TN	14	517
PAM50–6	mRNA	TCGA	TN	55	15	0.79	0.97	0.76
			non-TN	15	515

The values show the number of consistent and inconsistent samples clustered using each signature panel and identified using immunohistochemistry staining. F1 score that captures both false positives and false negatives is used to assess the classification accuracy. ‘TN’ and ‘non-TN’ are triple negative and non-triple negative breast cancers, respectively. PAM50 and PAM50–6 (with ER, PR, HER2, FOXA1, MYC, MYBL2 excluded) are used as a benchmark. ‘F1_TN’ and ‘F1_nonTN’ each refers to F1 scores in identifying TN and non-TN tumors

ISSN: 1471-2164