False Positive and False Negative Rates Sample Clauses

False Positive and False Negative Rates. The two control samples, Sample A and Sample B, allows us to evaluate the false positive and false negative rates associated with each of the fitted two-groups models. f0 ∼ Truncated Normal Mixture, f ∼ Normal Mixture Local fdr < 0.1 Local fdr < 0.051 HCL 226 (14.71%) 201 (13.09%) f0 ∼ Truncated Skew-Normal Mixture, f ∼ Skew-t Mixture Local fdr < 0.1 Local fdr < 0.043 HCL 105 (6.84%) 91 (5.92%) f0 ∼ Truncated Skew-Normal Mixture, f ∼ GH Local fdr < 0.1 Local fdr < 0.089 Table 3.8: Number and proportion of significant proteins under each of the fitted two-groups models. Since we know that all proteins in these two samples were mixed in a 1:1 ratio, all proteins that are declared as significant are by default, false positives. However, not all the identified proteins in these two data sets are yeast proteins. There are 9 identified contaminants in Sample A and 7 identified contaminants in Sample B. These contaminants are highly unlikely to be present in equal amounts in the light and heavy labeled samples. In fact, the log-2 relative expression ratios of these contaminants are all significantly different than zero. In this sense, the presence of these contaminants provides us a way to also calculate a false negative rate for our methods. I.e., we declare as a false negative any of the contaminant proteins that our methods fail to identify as being differentially expressed. However, since the number of false negatives are quite small, the estimates of the false negative rate are likely to be less reliable compared to the false positive rates. The results given in Table 3.9 demonstrates the improved performance of the two two-groups models fitted with asymmetric components: the skew normal - skew-t, and the skew normal - generalized hyperbolic combinations. With our data, the performance of the skew normal - generalized hyperbolic two-groups model, at least with respect to the false positive rate, is clearly superior to the other model setups considered. Sample A, m = 614 Sample B, m = 588 Proteins with locfdr < 0.1 56 39 21 47 31 16 False Positive Rate 8.26 % 5.12 % 2.15 % 7.23 % 4.48 % 1.89 % False Negative Rate 33.33 % 11.11 % 11.11 % 28.57 % 28.57 % 28.57 % Table 3.9: False positive and false negative rates for the three two-groups models. TNM-NM = Truncated Normal Mixture - Normal Mixture, TSNM-STM = Truncated Skew Normal Mixture - Skew-t Mixture, TSNM-GH = Truncated Skew Normal Mixture - Generalized Hyperbolic.