A clustering-based approach for efficient identification of microRNA combinatorial biomarkers

BMC Genomics

Table 5 Comparison of feature selection methods on two data sets^a

Methods	Feature #	GSE22220			GSE40525
		Sensitivity	Specificity	TA	Sensitivity	Specificity	TA
CFS	29/6	0.984	0.744	0.783	0.942	0.925	0.933
BFS	4/3	0.953	0.733	0.758	0.904	0.904	0.904
χ ²	2	0.976	0.701	0.729	0.923	0.906	0.913
	3	0.913	0.753	0.763	0.923	0.923	0.923
	4	0.945	0.764	0.787	0.923	0.923	0.923
Consistency	13/5	0.953	0.771	0.797	0.942	0.925	0.933
IG	2	0.976	0.701	0.729	0.827	0.896	0.865
	3	0.913	0.753	0.763	0.942	0.925	0.933
	4	0.945	0.764	0.787	0.923	0.906	0.913
RF	2	0.976	0.701	0.729	0.923	0.906	0.913
	3	0.913	0.734	0.744	0.942	0.891	0.913
	4	0.953	0.747	0.773	0.942	0.907	0.923
t-test	2	0.913	0.753	0.763	0.923	0.906	0.913
	3	0.890	0.807	0.802	0.923	0.889	0.904
	4	0.890	0.837	0.826	0.942	0.891	0.913
Wilcon test	2	0.913	0.753	0.763	0.923	0.906	0.913
	3	0.890	0.807	0.802	0.942	0.891	0.913
	4	0.937	0.793	0.812	0.942	0.925	0.933
CluFDA^b	2	0.969	0.750	0.783	0.923	0.923	0.923
	3	0.976	0.775	0.812	0.942	0.925	0.933
	4	0.906	0.833	0.831	0.962	0.943	0.952

^aThe numbers before and after ‘/’ denotes feature numbers of GSE22220 and GSE40525, respectively Sensitivity = TP/(TP+FN), Specificity = TN/(FP+TN) TA: total accuracy
^bCluFDA denotes the clustering-based feature selection using FDA method for selecting representative miRNAs

ISSN: 1471-2164