Application of Pearson correlation coefficient (PCC) and Kolmogorov-Smirnov distance (KSD) metrics to identify disease-specific biomarker genes

Table 1 Top 10 gene pairs for top prediction accuracies on PCA diagnosis.

Down gene	Up gene	True positive	True negative	Accuracy
PCC sort*
ACTA1	CRISP3	67/90	73/81	140/171
TGFB3	BICD1	72/90	68/81	140/171
ACTA1	HPN	76/90	63/81	139/171
MYL9	CRISP3	64/90	75/81	139/171
AL044599	BICD1	75/90	64/81	139/171
DMN	CRISP3	65/90	73/81	138/171
GJA1	CRISP3	70/90	68/81	138/171
AL036744	CRISP3	65/90	73/81	138/171
DMN	BICD1	69/90	69/81	138/171
ADH5	BICD1	71/90	67/81	138/171
KSD sort**
GSTP1	CRISP3	68/90	72/81	140/171
AOC3	CRISP3	69/90	70/81	139/171
GSTP1	UBE2C	66/90	73/81	139/171
HLA-E	RGS10	71/90	68/81	139/171
GSTP1	HPN	70/90	68/81	138/171
DMN	CRISP3	65/90	73/81	138/171
GJA1	CRISP3	70/90	68/81	138/171
HLA-E	UBE2C	61/90	77/81	138/171
DMN	BICD1	69/90	69/81	138/171
PALLD	BICD1	66/90	72/81	138/171

^*PCC sort: significant genes were separated into down- and up- regulated groups, then the top 20 genes (sorted by Pearson Correlation Coefficient in the cancer vs. normal GEPs for each gene) in each group were selected to generate pair-wise gene-pairs for the PCA prediction.
^**KSD sort: significant genes were separated into down- and up- regulated groups, then the top 20 genes (sorted by Kolmogorov-Smirnov Distance in the cancer vs. normal GEPs for each gene) in each group were selected to generate pair-wise gene-pairs for the PCA prediction.

ISSN: 1471-2105