I have experimental and computational data for n genes in data matrix. The rows are the genes and columns are its values. Now, the data is from two sources:

  1. Computational Prediction Results in form of p-values from Fisher Exact Test.
  2. Count Data results from the experiments across 4 different time points. This count data is basically is the number of signals you get per time point and it could be any positive number. There is no fixed bound.

I am interested in ranking the genes and therefore, I would like to know if there is a way to combine the computational prediction and experimental count result in the form of a statistic or some through some equation. I thought of sorting them in terms of p-values and then in terms of count data but I think it is very naive to do that way.

  • $\begingroup$ It's not clear what the ranking is to achieve. Why are you ranking? What will that get you? $\endgroup$ – Glen_b Oct 21 '13 at 20:28
  • $\begingroup$ Well ranking will help me to get the desired list of genes which I can use for further biological validation. $\endgroup$ – snape Oct 21 '13 at 21:57
  • $\begingroup$ What makes something more or less desirable for that purpose; why is p-value being used at all, rather than effect-size, say? $\endgroup$ – Glen_b Oct 21 '13 at 22:17
  • $\begingroup$ So, as I have mentioned in my questions I have 2 types of data. One from the computational result. P-value will shed some light on the gene that is getting enriched as compared to the other lists. This could be a false negative and therefore, I am using the experimental results to remove the false positives. So, the intersection of two groups will give me the list of genes that I am interested in. Now since, it is not feasible to do the biological validation for every genes; I am interested to rank the genes based on both count data and p-values. $\endgroup$ – snape Oct 21 '13 at 22:32
  • $\begingroup$ You're giving absolutely no basis on which we might suggest how to approach a tradeoff. $\endgroup$ – Glen_b Oct 21 '13 at 22:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.