Question: (Closed) How do I validate my findings?
gravatar for Na Sed
4.8 years ago by
Na Sed300
United States
Na Sed300 wrote:

I am working on effective miRNA-mRNAs interactions in cancer A. So I just do statistical tests for tumor samples.

Now, I have been asked to validate my findings, claimed effective miRNA-mRNA interactions, using Normal samples. However, I do not have normal samples!

I remember that I saw somewhere or heard from someone that in such case which I do not have normal samples, I can consider an unrelated cancer and do the same tests on this cancer. Then, I can compare the results of both of them and count how many interactions are common between two cancers. If there is no common interactions, or just a few common interactions, I can say my findings are significant and reliable.

Now I have three questions:

1) What do you think about the bolded paragraph? Is it correct?

2) Is it rational if I run my tests on a related cancer and expect to have many common miRNA-mRNA interactions? If so, how many interactions must be common to say that my results are reliable and significant?

3) Is it rational if I run my tests on normal samples from a different tissue?  In this case, must I do expect to have no common interactions between the cancer and these normal samples?

PS. If you know some references in this context, I appreciated if you introduce to me.

ADD COMMENTlink modified 8 months ago by Biostar ♦♦ 20 • written 4.8 years ago by Na Sed300

I'm not sure that Biostars is the right place.

Can you define what you mean by "do statistical tests for tumor samples"? Are you using miRNA-Seq to quantify miRNA expression? Are you predicting miRNAs from genomic data?

However, assuming you're thinking that some miRNA has an anti-tumor effect: I think you might be confusing two different things, validating the target(s) of the miRNA versus demonstrating the miRNA's effect on its target(s) imparts an anti-tumor effect.

Validating what the miRNA binds doesn't require healthy tissue but showing it has a specific antitumor and not a generic cytotoxic effect (i.e. kills tumors and not any cells) is much more complicated.

I do not think what you propose in the bolded paragraph makes any sense.

For one, barring mutations, the miRNA will presumably have the same possible targets in any cell type in that species. The only difference will be the expression level of those targets and their biology. Second, I'm not sure how showing that the expression of your miRNA's target(s) are absent from another tumor tells you anything about the anti-tumor effect of your miRNA.

Even if the target(s) are present in both tumor types, that doesn't tell you that your findings are wrong either.Imagine a case where a miRNA targets some mRNA expressed in two different tumor types. In one tumor the protein encoded by the mRNA is required for survival. In the other tumor the expression of that factor is not required for survival. So, your miRNA would work in one tumor but not another, but does not tell you if your miRNA will kill non-cancerous tissue of the susceptible tumor.

Running the test in healthy tissue is the same thing, especially if in different tissue types.

If you're talking about seeing that the miRNA of interest has some non-zero expression (i.e. is expressed) is much different than seeing if it is deferentially expressed compared to normal tissue. You could be seeing the expression of a miRNA that is expressed in both cancerous and healthy tissues.

ADD REPLYlink modified 9 months ago by RamRS30k • written 4.8 years ago by pld4.8k

Thanks for your answer.

Indeed, I am predicting targets for miRNAs using their expression profiles by GenMiR++ method. So I obtain a matrix which rows are mRNAs and columns are miRNAs. If entry [i,j] is non-zero, I can say miRNA j targets mRNA i; in the other words there is an interaction between miRNA j and mRNA i.

After obtaining such matrix, I just consider top-ranked 1000 interactions and claim that they might be effective in developing the correspondence cancer. Now, I have been asked to run such procedure on normal samples and find top-ranked 1000 entry[i,j] which are non-zero to compare the obtained interactions in tumor and normal samples, while I don't have normal samples!!

ADD REPLYlink modified 9 months ago by RamRS30k • written 4.8 years ago by Na Sed300

I do not thing you can make that claim, those miRNAs could be expressed at the same level in healthy tissue. I honestly think the only claim you can make is that, in your particular samples, miRNAs x,y,z are expressed and potentially target a,b,c. That is it, without the proper controls, there is no way to substantiate any claims as to their role (or lack of) in cancer.

Do you have any idea of outcomes (survivors, treatment and response, etc)? Maybe you can see if there is differential expression of miRNAs between tumor classes. However, you'll have to be very careful to make sure you're not confounding meaningful comparisons with other factors.

You might want to check around on databases, or see if you can't find a way to obtain tissues, there are a few programs out there that exist to provide human tissues for scientific research. Obviously you will run into issues comparing libraries/sequencing on different runs, but it is better than nothing.

The takeaway lesson here is to always have controls, always. Controls are everything. Experiments without controls are a waste of time and funding.

ADD REPLYlink written 4.8 years ago by pld4.8k

Old question, closed to avoid bumping by the Biostars bot.

For this reason we have closed your question. This allows us to keep the site focused on the topics that the community can help with.

If you disagree please tell us why in a reply below, we'll be happy to talk about it.


ADD REPLYlink modified 8 months ago • written 8 months ago by ATpoint38k
Please log in to add an answer.
The thread is closed. No new answers may be added.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 801 users visited in the last hour