Inter-platform concordance of gene expression data for the prediction of chemical mode of action

Biology Direct

Table 4 Accuracies of predicting MOA’s in the originally given test set, based on classifiers developed on complete gene expression sets profiled from microarray and RNASeq platforms

Platform	Classifier	Overall Acc. %	Sensitivity, Specificity
			PPARA	CAR/PXR	Control	OTHER
Microarray	Ensemble	55	33,61	44,58	100,48	56,54
	SVM	55	0,70	0,70	83,50	100,21
	RF	55	0,70	0,70	83,50	100,21
	PLS+LDA	38	67,30	44,36	100,28	0,66
	PLS+RF	60	11,73	0,76	100,53	100,30
	PCA+LDA	17	0,22	11,19	83,6	6,25
	PCA+RF	38	0,48	0,48	0,44	89,0
	RPART	48	100,34	0,61	83,42	33,59
RNASeq	Ensemble	55	44,58	33,61	100,48	56,54
	SVM	55	0,70	0,70	83,50	100,21
	RF	52	0,66	0,66	83,47	94,20
	PLS+LDA	33	44,30	44,30	100,22	0,58
	PLS+RF	60	0,76	22,70	100,53	94,34
	PCA+LDA	10	0,13	22,7	33,6	0,18
	PCA+RF	43	0,55	0,55	17,47	94,5
	RPART	24	44,19	22,25	67,17	0,42

ISSN: 1745-6150