A longitudinal feature selection method identifies relevant genes to distinguish complicated injury and uncomplicated injury over time

Table 1 Performance of SAMGSR extension and other relevant algorithms on the injury data

Method	# of genes	Using 5-fold CVs				On the test set
		Error	GBS	BCM	AUPR	Error	GBS	BCM	AUPR
L-SAMGSR¹	97	0.442	0.268	0.515	0.576	0.356	0.230	0.535	0.622
EDGE¹	1083	0.442	0.281	0.511	0.526	0.407	0.234	0.514	0.594
SAMGSR separately^a	> 400	0.419	0.246	0.510	0.559	0.428	0.243	0.511	0.553
P-SVM separately	> 1000	0.488	0.281	0.477	0.454	0.441	0.244	0.511	0.560
LASSO separately	147	0.465	0.261	0.497	0.498	0.407	0.237	0.509	0.580

Note: ^a the posterior probabilities were calculated using an SVM classifier. Here, the cutoff for q-value in SAM-GS part is set at 0.05. # of genes represents the number of the union of individual genes selected at each time point. CV: cross-validation

ISSN: 1472-6947