Figure 4

Logistic regressions of the probability of detection of gene expression differences from simulated data. Logistic regressions of the frequency of affirmative significance call over log₂ factor of difference in gene expression. The logistic model plotted is that log_e(p/(1 - p)) = mx + b, where x is the log₂ factor of difference in gene expression. Cross symbols represent actual data points. Each is placed at its estimated expression level, either at the top of the plot. When identified as significant (S), or at the bottom when identified as not significant(NS). Logistic regressions are of statistical significance calls A) on the "true" factors of fold change from which data was simulated. The model has a highly significant fit (χ² = 884.5, P < 0.0001). The estimated intercept for the log odds, b, of an affirmative significance call is -16.4 (significant, P < 0.0001). This corresponds to a probability of a positive call of 0.02, which is the observed average false-positive rate. The estimated slope with log₂ factor of difference in gene expression, m, is 12.5 (significant, P < 0.0001). B) on the factors of difference estimated from the simulated data. The model has a highly significant fit (χ² = 890.5, P < 0.0001). The estimated intercept for the log odds, b, of a significant call versus no significant call is -3.9 (significant, P < 0.0001), and the estimated slope with log2 factor of difference in gene expression, m, is 10.7 (significant, P < 0.0001).