The figure shows the results of analysis of artificial data-sets. Throughout means ± SEM of the 40 re-samplings are given. Top panel: This panel shows the variance reduction by normalization and how this depends on the bias-to-variation ratio. Increasing the bias-to-variation ratio in the artificial data leads to more variance being removed from the data-set by NORMA-Gene. Middle and bottom panels: These panels show the effectiveness of NORMA-Gene versus reference gene normalization and the dependence of different parameters. For each of the re-samplings (see methods for detailed information on the construction of the data-sets) the proportion of NORMA-Gene normalized data points that were closer to the true mean was calculated and is shown on the y-axis (0.5 represent equal performance of NORMA-Gene and reference gene normalization). Middle panel: Here the x-axis represents the ratio between bias (between replicate variation) and variation (between genes within replicate variation). As bias among replicates increase the performance of NORMA-Gene decreases. No qualitatively difference was observed in the performance of NORMA-Gene between re-sampling from eight or four genes. Bottom panel: The x-axis represents the ratio of reference gene-to-target gene variation (here 1 represent equal variation and 0.25 represent four-fold decreased variation in the reference factor).
Heckmann et al. BMC Bioinformatics 2011 12:250 doi:10.1186/1471-2105-12-250