hi .. i think is better to use fold changes to determine the differences between control healthy  peoples and patients 

you can use graphpad prism to determine the fold changes and then to determine ROC

best of luck

I would suggest the fold-changes too.  Raw Ct values won't give you anything since it is not normalized to the reference gene.  Also, how do you define "normalized Ct value"?  Is it normalized to the reference gene?

For a single miRNA, the normalized Ct values are easily obtained from a set of participants including patients and healthy people.

Comparably, application of fold change may not take the extensionality of the ROC model into account. Briefly, it may be difficut to calculate the fold change of the miRNA for a new participant outside of the set.

For details, you can see my article titled "Neural Network Cascade Optimizes MicroRNA Biomarker Selection for Nasopharyngeal Cancer Prognosis".

Article Neural Network Cascade Optimizes MicroRNA Biomarker Selectio...

Many thanks Zainab,  Shuai and  Wenliang  for your reply, however I have the same argument as Wenliang. Clinically (in the future if our microRNA succeeded as a biomarker) we will ask for its level in the patient only, we cannot ask for a healthy control blood every time we do the test to obtain a fold change.  So may be the normalized ct will give us more accurate sensitivity and specificity measures

 Thank you for clarifying your situation!

Presumably by "normalised Ct" you mean delta Ct, the difference between your miRNA of interest and a reference miRNA within a sample? 

If so, you should be able to use this on a ROC curve. Calculation of this does not require comparison to a control, like fold change does.

Thanks Christopher, that's exactly what I mean by normalized ct (delta ct). and thanks for your answer

Since it is a patient study therefore delta Ct ( Target miR-Ref miR) is to be calculated and not by delta delta ct.  Mean Ct value of all patients vs mean ct value of all healthy can then be compared only for determining the fold change. Otherwise Median,  Quartiles of ct values can then be analyzed for  ROC. If  no direct correlation is obtainable (assuming human data is every where) and there are multiple variables together with miRs  then Principal component analysis is also an interesting way of interpretation.

Before you plot the ROC, which plot shows a greater difference/separation between disease and normal? (plot of the fold change, or plot of the Ct threshold derivative?)

Dear Louis:

many thanks for your reply, right now we are plotting the fold changes only, we never tried to plot the normalized Ct, but I think they will give a similar results as calculating the fold change depends on the values of normalized Ct in the experimental groups. 

any suggestion??

please am just a starter in miRNAs expression analysis, please can someone tell me how to prepare my data to plot a ROCcurve. l have done the expression analysis using  DESeq2 in r but am having difficulty on how to plot the ROC. please l need help

A ROC curve is an approach that will be used to evaluate classification and prediction of models in terms of sensitivity and specificity (95% CI)  to draw conclusions, for patients based data you have to use relative miRNA expression. If it was measured patients tissue or relative levels if it is circulating miRNA levels from plasma/serum ( In that case normalized normalized Ct value subtracted from  Ct value of exogenous spike in control that (if) you have added) for that sample (power 2(-delta Ct) will be normalized for that individual patient ΔCt=Ct target miRNA – Ct spike-in miRNA) . Some studies have used average of spike in control CT as one value and all other patients miRNA values are subtracted from this one value of ("N'patients, where N=1,2....n) The values that you will obtained will be your normalized values and they will be used as data point versus your phenotype for determining ROC. Fold change for this kind of study is only possible when an average of all your healthy controls normalized delta Cts compared to the average of all your patients delta Cts. Check the following    Schmittgen TD, Livak KJ (2008) Analyzing real-time PCR data by the comparative C(T) method. Nat Protoc, 3:1101-8.