I've used GraphPad Prism to draw the data from REST and these graphs can be directly used for publications. Choose in Column Graph : Box and Whiskers : 2.5 - 97.5 percentile. Add your data in the columns: minimum - min error bar- median-median-median-max error bar- maximum (http://www.graphpad.com/guides/prism/7/user-guide/index.htm?box-and-whiskers.htm , last paragraph). In graph: format y-axis by double clicking on it. Choose for log 2 scale and determine your minimum and maximum range of axis. Choose for the 'antilog' function in number format. 

In this way you obtain the same graphs as in REST. I hope this will help! 

Yes, you can use means of the triplicates in "Rest".

About the 'control vs treatment', it depends on the design of your experiment.... Usually yes, we compare the effect of a treatment against a non treated sample.

You can show graphs from "Rest" analysis which present the errors bars. The output tells you if you Cps mean DOWN or UP regulated.

It's up to you if you only want to show a table with p values. I believe it's enough information.

If you did the PCRs as a master mix divided by two, one half used primers for normalizer gene and the other half with primers for target gene, I would say normalize each replicate first, then you do the MNE.

I hope it helps. Cheers, Rute

I personnally put the raw data in directly rather than the means from a triplicate. We have a rotorgene so it is simple to paste the raw data in. The problem comes if you have more than one treatment group. Then you produce a number of rest files for each treatment versus the control. To produce nice graphs then requires some fiddling. You can use the dat in the table to make whisker box plots in excel (search using google). I often then paste the figures from Excel ino a vector drwawing pacakage to clean them up for producing hi resolution images. It takes some time but given REST is free you can't really complain. Also REST produces a more statistically robust p value.

Thanks for the advice. I did do a master mix and split it for target and normalizer. I'm assuming you just randomly choose which wells to pair together when normalizing the replicates first.

I do have the problem of more than one treatment group. I have 2 treatments, 3 doses each, and 3 time points. So, I did have to run REST separately for each group. I did find how to do box and whisker plot in excel, but not specifically for using the data REST gives you. I'm thinking its the way I'm using the numbers that my plots aren't the same as displayed in REST. I used the standard error for Q1 and Q3 and the CI for min and max, but I suppose that's not correct now that I think about it. Is there a way to graph in excel using the REST values without having to use excel first to calculate, min, max, median, etc. or maybe I should just use some statistical software?

you don't really need to calculate anything when using REST, just feed the software with the raw Ct values from the qPCR device and specify what they are (reference gene, target gene, control sample replicate and so on), REST does the rest. That's the poinof using it inead of Excell. It also gives you the graphs in the end that you can use straight for the publication.

Be carefull with the graphs. When you use the mean value for bars and the standard error for error bars sometimes this data are presented (visually) as no difference because the error bars can show values that are at the same point that the control bar. The up-regulation or down-regulation of your treatment sample is signifficant when the p values are under 0.05. Some editors can question you about that.

This has been very helpful. Particularly the explanation of 1 being the control and then up or down regulation from that point.

I can't directly use the graph from REST, because of my multiple treatments so I would like to combine the multiple treatments onto one graph. Since Excel doesn't do box and whisker plots its a bit of challenge. I haven't been able to get the upper error bar to be the correct length so I may have to play with it some more.

I can see what you're saying with the error bars. I do have some samples that show significant down-regulation, but with the upper standard error it would be greater than 1 even though the p value is less than 0.05.

The graph present two informations. One, the mean value of expression with the error, and second, the p value when the difference is significant. Most of time, but not always, these informations are complementary. Remember that the p values are show as result of values comparison (control vs. treatment) and are not a result of statistical analysis.

I'm confused, the p value isn't a result of the statistical test? I thought the p-value was the result of running the Pair Wise Fixed Reallocation Randomisation Test. What's the point of the p-value if it isn't showing statistical significance?

I was finally able to duplicate the graphs from REST into excel, so I can present the data in the same box and whisker manner.

REST provides a statistically valid p value through randomization (~permutation) of the raw data (samples/treated and controls/untreated) and calculating the expression ratios of the randomized/permuted data. Usually it performs =>10000 randomizations, so that the set is well distributed. Then, the test checks the percentage of those randomized expression ratios that are higher than the original non-randomized (i.e., real) results. If it is less than 5% (=p

Ok, that is how I understood it.

On an unrelated subject of boxplots: you should consider using the R statistics environment (http://www.r-project.org/) or the more costly S for your statistics and data plotting. A bit harder, but much more powerful than Excel.

Hello,

I have the same problem as you Kayla, I have multiple treatments in my experiment and would like to show the whisker box plots for the different treatments on the same graph, can you tell me how you did it?? thank you in advance

Hi Sarah,

I ended up switching methods entirely and not using REST. I switched to using delta-delta Ct (Livak method) and generating bar graphs. It was a little more work but once I had all of my data in excel it was easy to enter formulas to do all of the calculations and error bars. I used graphpad to generate my graphs which gave me more control on the appearance on the graphs than excel does.

When presenting multiple REST data I use the mean, quartile and 95% CL values from the table to plot boxplots in exel using a cheaty method

http://support.microsoft.com/kb/155130

This works quite well and if I am being really anal I then copy them into a vector drawing package to clean them up and make them look nice.

thanks for your answer Kayla! Noel, my problem is actually that I don't know which values generated by REST I should use. The REST output shows the mean of the expression ratios, the standard error (68% confidence interval) and the 95% confidence interval and I don't know how to use these values to calculate the quartiles and the median, can you help me with that?? thank you very much

Hi 

Can anyone explain why in REST, the p value does not correspond to the relative expression. 

For example: In one of my tests if I have an expression value of -9 and the p value is not significant and then in another test the expression value is -10 and the p value is significant. Surely the greater the difference the more likely it is to be statistically different? 

Thanks

Catríona 

Hi Kayla, could you please explain how you duplicated graphs from REST to excel?

I've used GraphPad Prism to draw the data from REST and these graphs can be directly used for publications. Choose in Column Graph : Box and Whiskers : 2.5 - 97.5 percentile. Add your data in the columns: minimum - min error bar- median-median-median-max error bar- maximum (http://www.graphpad.com/guides/prism/7/user-guide/index.htm?box-and-whiskers.htm , last paragraph). In graph: format y-axis by double clicking on it. Choose for log 2 scale and determine your minimum and maximum range of axis. Choose for the 'antilog' function in number format. 

In this way you obtain the same graphs as in REST. I hope this will help!