A comparative study of discriminating human heart failure etiology using gene expression profiles

BMC Bioinformatics

Table 11 Percentiles of misclassification errors from 50 simulated datasets for two-class classification. The true model is either PPLS or LASSO fitted with top 400 genes to the Minnesota data to compare ischemic vs idiopathic.

Methods	# of top genes	True model: PPLS						True model: LASSO
		Mean	0%	25%	50%	75%	100%	Mean	0%	25%	50%	75%	100%
PPLS	50	2.54	0	2	2	3	6	0	0	0	0	0	0
	100	2.6	1	1.25	2	3	7	1	1	1	1	1	1
	400	2.58	0	2	2	3	7	0	0	0	0	0	0
	1600	2.62	0	2	2	3.75	7	0	0	0	0	0	0
	6400	2.6	0	2	2	3	6	0	0	0	0	0	0
PLS	50	2.58	0	2	2	3	6	0	0	0	0	0	0
	100	2.6	0	2	2	3	7	0	0	0	0	0	0
	400	2.56	0	2	2	3	7	0	0	0	0	0	0
	1600	2.52	0	2	2	3	7	0	0	0	0	0	0
	6400	2.54	0	2	2	3	6	0	0	0	0	0	0
LASSO	50	2.48	0	2	2	3	6	0	0	0	0	0	0
	100	2.48	0	2	2	3	6	0	0	0	0	0	0
	400	2.84	0	2	2	3	10	0	0	0	0	0	0
	1600	3.48	0	2	2	4.75	10	0	0	0	0	0	0
	6400	3.48	0	2	2	4.75	10	0	0	0	0	0	0
SC	50	2.76	1	2	2	3	7	1	1	1	1	1	1
	100	2.74	1	2	2	3	7	1	1	1	1	1	1
	400	2.96	0	2	3	3.75	7	0.28	0	0	0	1	1
	1600	3.58	0	2	3	5	8	0	0	0	0	0	0
	6400	4.08	1	3	4	5	10	1	1	1	1	1	1
RF	50	2.48	0	2	2	3	6	0	0	0	0	0	0
	100	2.48	0	2	2	3	6	0	0	0	0	0	0
	400	2.48	0	2	2	3	6	0	0	0	0	0	0
	1600	2.48	0	2	2	3	6	0	0	0	0	0	0
	6400	2.48	0	2	2	3	6	0	0	0	0	0	0

ISSN: 1471-2105