next up previous contents
Next: 3.3.11.3 Exporting the results Up: 3.3.11 Correlation Analysis Previous: 3.3.11.1 Calculating correlations   Contents

3.3.11.2 Visualizing results

The last parameter in Figure 3.67 allows to choose which results should be displayed. The choice may be made before the analysis is launched, but it can also be modified (without further calculations) after the analysis, from the menu Tools $\rightarrow $ Correlation Analysis $\rightarrow $ Redisplay:

Figure 3.68: Tools $\rightarrow $ Correlation Analysis $\rightarrow $ Redisplay - Correlation Analysis: Redisplay dialog.
Image gtca-redisplay-dialog

The Redisplay dialog proposes several options:

correlation coefficient:
(default) Pearson or Spearman correlation coefficient, depending on the choice made within the Compute dialog
p-values:
significance level assigned to each correlation coefficient
FWER:
p-values adjusted for multiple comparison in order to control the Family-Wise Error Rate (FWER), that is, the probability that one or more loci among those selected is a false positive. We use the Holm adjustment procedure, which is more powerful that the traditional Bonferroni procedure
FDR:
p-values adjusted for multiple comparison in order to control the False Discovery Rate (FDR), that is, the expected proportion of false positives among those loci selected

Figure 3.69: Visualizations of the results of Correlation Analysis. From top to bottom: correlation coefficients; p-values; FWER-adjusted p-values; FDR-adjusted p-values. All p-values are plotted in $-log10$ scale, and given the sign of the corresponding correlation coefficient.
Image gtca-corr-panel-info Image gtca-pvalues Image gtca-fwer Image gtca-fdr

These FWER or FDR multiple testing adjustments may be performed for the whole set of genes (``whole genome''), or for each chromosome separately (``by chromsome''), leading to a less conservative adjustment. If only one chromosome is selected, choosing ``whole genome'' or ``by chromosome'' therefore gives exactly the same results.

Importantly, all (adjusted and unadjusted p-values) are displayed in $-10\log$-scale, so that significant genes (with small p-values) can be easily seen. For example, a gene with p-value $10^-7$ will be plotted with $y=7$. Also note that all p-values are given the sign of the corresponding correlation coefficient (which may be negative for some genes).


next up previous contents
Next: 3.3.11.3 Exporting the results Up: 3.3.11 Correlation Analysis Previous: 3.3.11.1 Calculating correlations   Contents
2007 - Institut Curie Bioinformatics unit