Cytokine expression profiles of immune imbalance in post-mononucleosis chronic fatigue

Table 2 Performance Characteristics ¹ of Subject Classification using Cytokine Combinations

	Optimal Linear model subset		Top ranking AUC with CC correction		Top ranking U statistic with CC correction
	Classification based on IL-2, 6, 8, 23, IFNg		Classification based on IL-2, 5, 8, 13, 23		Classification based on IL-5, 23, 8, IFNg, TNFb
	Linear cw 90% Conf	Linear uncorrected	Linear cw 90% Conf	Linear uncorrected	Linear cw 90% Conf	Linear uncorrected
Correct Rate	0.97	0.90	0.94	0.86	0.92	0.79
Error Rate	0.03	0.10	0.06	0.14	0.08	0.21
Inconclusive Rate	0.17	0.00	0.62	0.00	0.69	0.00
Classified Rate	0.83	1.00	0.38	1.00	0.31	1.00
Sensitivity	0.72	0.94	0.33	0.83	0.33	0.67
Specificity	0.88	0.88	0.38	0.88	0.25	0.88
Positive Predictive Value	0.81	0.85	0.29	0.83	0.25	0.80
Negative Predictive Value	0.81	0.95	0.43	0.88	0.33	0.78
Positive Likelihood	5.78	7.56	0.53	6.67	0.44	5.33
Negative Likelihood	0.32	0.06	1.78	0.19	2.67	0.38

Classification models based on random all-possible-subset selection of 5 cytokines as well as the top 5 ranking cytokines based on individual contribution to the AUC and the U statistic.
¹ Correct Rate is defined as (correctly classified samples)/ (all classified samples); Error rate as (incorrectly classified samples)/ (all classified samples); Inconclusive Rate is defined as (non-classified samples) / (total number of samples); Classified rate is (classified samples) / (total number of samples); Sensitivity is defined as (correctly classified positives) / (true positives); Specificity is (correctly classified negatives) / (true negatives); Positive Predictive Value is (correctly classified positives) / (positive classified); Negative Predictive Value is (correctly classified negatives) / (negative classified); Positive Likelihood is Sensitivity / (1 – Specificity); Negative Likelihood is (1 – Sensitivity) / Specificity.

ISSN: 1479-5876