Hello,
I am having micro RNA data sets of control and patients now I am interested into analyzing micro RNA as a biomarker for a particular disease. I have already delta delta CT values for miRNA expression. In order to do so I want to develop ROC curve, if possible would you please explain step wise how to calculate ROC and AUC/ what statistical software you used for analysis?. This will be very helpful for me in my future research.
Thanking you in advance.
Mayur Doke
Dear Dr. Mayur Doke,
In my experience using microarray data and RNA-Seq data I was able to use the ROC curve based on a strength prediction of my Bayesian network data. For example, if I were using a Bayesian network and attempting to calculate a Roc curve from the data I would utilize the R Package bnlearn. Here were my steps:
1st) Calculate a Bayesian Network in bnlearn using the hill climbing (hc) greedy search code and a data matrix of your microRNA data that is discretized while the scoring you would like to use is bde.
Example code: j = hc(PMID3, R = 200, m =30, score = "bde")
2nd) Model your network titled "j" into a true directed acyclic graph.
Example code: true.dag = model2network (j)
3rd) Calculate the strength of your network from step 2 using the boot.strength code and hill-climbing algorithm (this allows your to perform the prediction of the true positive and false positive rates used to graph your ROC curve). R stands for repetitions your model was run in the algorithm to determine the positive and negative result rates.
strength1 = boot.strength (true.dag (Here is where you put in your mRNA data matrx) , R = 200, m =30, algorithm = "hc")
4th) Perform prediction in bnlearn. In this step you calculate the predictive rates compiled from the strength code in step 3 and also from the BN created in step 1.
Example Code: pred = as. prediction(strength1, j)
5th) In this step you calculate the performance of your predictive model in step 4 also calculating the scores for true positive rate "tpr" and false positive rate "fpr" used in the next step.
Example Code: perf = performance(pred, "tpr", "fpr")
6th) Plot the ROC curve and calculate the area under the curve. After entering this code your ROC curve should appear on the plot panel of R.
Example Code: plot(perf, main = "Arc Detection")
7th) Calculate the area under the curve score for model validation. In this final step you will generate the AUC score that explains the strength of your model and how best your results are predicted accurately.
Example Code: performance(pred, "auc")
Other R packages that can be useful for microRNA data would be pROC and ROCR
I've attached a screenshot of the code I generated for a ROC curve using bnlearn and the ROC curve plot that was also generated. If you need any further help please do not hesitate to ask me.
My Best,
Christian
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software