Data Presentation

[Fancy Nancy]

I used ANOVA to analyze my log2 transformed relative quantities to assess statistical significance between 5 treatments and a control. What is the best way to graphically present my data. Do I graph the log2 transformed data (fold changes) or the relative quantities? What type of error bars do I use and how do I calculate the error? Is error propagation important?

I followed the following path to get to the fold changes:
1) Correct for efficiency
2) interplate calibrators
3) QPCR repeats
4) reference genes (two ref genes)
5) relative quantities
6) log2

Thank you!

[Mikael Kubista]

The pre-processing scheme is appropriate.

Typically expression data should be compared in logarithmic scale, because they tend to be (more) normal distributed in log scale. In GenEx Descriptive Statistics, if you remove the tick in the box “Plot only” GenEx will perform a normality data of your data. If most groups/genes pass you can go ahead with analysis.

Typically data are presented as bar graph with confidence intervals (use descriptive statistics). Confidence intervals are most appropriate because the degree of overlaps between groups indicate graphically whether the means are likely to be different are not. You should also report the p-values of the post test, at least for those genes selected for comparison that show significant differences.

Analyzing data in logarithmic scale error propagation is automatically accounted for. When Cq values are added or subtracted during pre-processing error contributions automatically add up. As consequence, unnecessary pre-processing should be avoided, since it increases the error. Any error in the estimate of PCR efficiency is not accounted for. The PCR efficiency does NOT influence the statistics (p-values) when comparing groups (there is no effect at all if the efficiency of the reference and reporter genes are the same and a negligible effect if they are different). This is because means and confidence intervals scale the same with PCR efficiency and two groups are differentially expressed no matter the PCR efficiency! PCR efficiencies are only important to estimate fold changes; i.e., the degree of differential expression.

Good luck

[Fancy Nancy]

Thank you for the explanation. I have found the error propagation issue to be quite confusing since various publications and software appear to have various opinions on its use. Your reply is very helpful.

[cDiaz]

Thank you Mikael Kubista very much for your post. It helped me to understand everything better. Great job!