Analyzing gene expression data for pediatric and adult cancer diagnosis using logic learning machine and standard supervised methods

Table 3 Analysis of gene expression profiles in eight selected data sets of for cancer diagnosis. Comparison between five methods of supervised data mining in cross-validation

Method	Sens. %	Spec. %	Youden Index %	Empirical Accuracy %	Cohen’s Kappa %	p
GDS4968
LLM	98.1	90.2	88.4	94.7	91.7	< 0.001
DT	96.2	95.1	93.1	95.7	93.3	< 0.001
ANN	94.3	95.1	89.4	93.6	90.0	< 0.001
SVM	98.1	97.5	95.7	97.9	96.7	< 0.001
kNN	98.1	97.6	95.7	96.8	95.1	< 0.001
GDS4887
LLM	100	95.0	95.0	97.5	95.0	< 0.001
DT	100	95.0	95.0	97.5	95.0	< 0.001
ANN	100	100	100	100	100	< 0.001
SVM	100	100	100	100	100	< 0.001
kNN	100	100	100	100	100	< 0.001
GDS4794
LLM	100	100	100	100	100	< 0.001
DT	100	100	100	100	100	< 0.001
ANN	100	100	100	100	100	< 0.001
SVM	100	100	100	100	100	< 0.001
kNN	100	100	100	100	100	< 0.001
GDS4762
LLM	100	100	100	100	100	< 0.001
DT	100	100	100	100	100	< 0.001
ANN	97.3	100	97.3	98.8	97.5	< 0.001
SVM	100	100	100	100	100	< 0.001
kNN	100	100	100	100	100	< 0.001
GDS4471
LLM	99.0	96.0	95.0	97.4	94.0	< 0.001
DT	88.2	76.0	64.2	84.2	64.2	< 0.001
ANN	82.4	88.0	70.4	84.2	66.3	< 0.001
SVM	98.0	96.0	94.0	97.4	97.4	< 0.001
kNN	94.2	96.0	90.1	94.7	88.3	< 0.001
GDS4296
LLM	97.8	96.2	94.0	96.6	95.7	< 0.001
DT	75.8	100	75.8	63.3	53.1	< 0.001
ANN	98.9	96.2	95.1	93.2	91.4	< 0.001
SVM	100	100	100	100	100	< 0.001
kNN	100	100	100	100	100	< 0.001
GDS3952
LLM	96.4	90.3	87.6	92.2	89.4	< 0.001
DT	94.5	100	94.5	70.2	57.7	< 0.001
ANN	97.3	90.3	87.6	76.6	67.2	< 0.001
SVM	100	100	100	95.7	94.2	< 0.001
kNN	100	100	100	97.2	96.1	< 0.001
GDS3945
LLM	100	100	100	100	100	< 0.001
DT	90.5	100	90.5	95.2	90.5	< 0.001
ANN	14.3	81.0	−4.8	46.6	−4.8	0.661
SVM	95.2	100	100	97.6	95.2	< 0.001
kNN	85.7	95.2	81.0	90.5	81.0	< 0.001

ISSN: 1471-2105