search

BridgesLab / CushingAcromegalyStudy

The source code for the cushing and acromegaly studies, currently ongoing
Other
1 stars 2 forks source link

Check if age is a confounding factor between methylation and expression #30

Closed qtran1 closed 9 years ago

davebridges commented 9 years ago

Which table should I be looking at for the acromegaly age adjusted stuff, the normal one or the LRT one. I am probably going to need you to write that particular methods section.

qtran1 commented 9 years ago

You should look at the normal table. It has all the comparisons. Yes, I can write that part.

davebridges commented 9 years ago

what do you think should be a reasonable cutoff for if we mention that there is an age dependent effect? I am starting by going through the genes we already discussed (q<0.05). I was mentioning the age as a factor if p>0.05 for one of the subgroups or if the age adjusted p-value was >0.05. I am thinking that might be too much, but id like your thoughts

On Tue, Sep 23, 2014 at 9:49 AM, Quynh Tran notifications@github.com wrote:

You should look at the normal table. It has all the comparisons. Yes, I can write that part.

— Reply to this email directly or view it on GitHub https://github.com/BridgesLab/CushingAcromegalyStudy/issues/30#issuecomment-56531571 .

qtran1 commented 9 years ago

I'm guessing that what you try to say is what genes are affected by age and disease, right? If that the case, it's easier to look at the LRT table. In this table, all the genes that have padj > 0.05 means that they may affected by gender or not at all and those with padj <0.05 means that they are affected by both age and disease.

Remember that, the results of all the test are not additive. That means the union of sets of genes in the Wald tests (pair-wise comparisons) is not expected to be the same as sets of genes in the f LRT test.

On Sep 23, 2014, at 9:58 AM, Dave Bridges wrote:

what do you think should be a reasonable cutoff for if we mention that there is an age dependent effect? I am starting by going through the genes we already discussed (q<0.05). I was mentioning the age as a factor if p>0.05 for one of the subgroups or if the age adjusted p-value was >0.05. I am thinking that might be too much, but id like your thoughts

On Tue, Sep 23, 2014 at 9:49 AM, Quynh Tran notifications@github.com<mailto:notifications@github.com> wrote:

You should look at the normal table. It has all the comparisons. Yes, I can write that part.

— Reply to this email directly or view it on GitHub https://github.com/BridgesLab/CushingAcromegalyStudy/issues/30#issuecomment-56531571 .

— Reply to this email directly or view it on GitHubhttps://github.com/BridgesLab/CushingAcromegalyStudy/issues/30#issuecomment-56533098.

davebridges commented 9 years ago

I guess what i want is to go through the genes we discussed and remove the ones that were confounded by age. We can talk about this more on thursday, but what i did was look at each gene we were interested in and checked if its padj<0.05 , then if that was true for each of the groups. I flagged the ones where the age adjusted p-values were >0.05 (to remove, since those are probably confounders) and then specified each case in which one group or the other was significant. If both groups were significant, i just added the age adjusted q-value/fold change. A lot of the genes we talked about disappear this way.

On Tue, Sep 23, 2014 at 10:29 AM, Quynh Tran notifications@github.com wrote:

I'm guessing that what you try to say is what genes are affected by age and disease, right? If that the case, it's easier to look at the LRT table. In this table, all the genes that have padj > 0.05 means that they may affected by gender or not at all and those with padj <0.05 means that they are affected by both age and disease.

Remember that, the results of all the test are not additive. That means the union of sets of genes in the Wald tests (pair-wise comparisons) is not expected to be the same as sets of genes in the f LRT test.

On Sep 23, 2014, at 9:58 AM, Dave Bridges wrote:

what do you think should be a reasonable cutoff for if we mention that there is an age dependent effect? I am starting by going through the genes we already discussed (q<0.05). I was mentioning the age as a factor if p>0.05 for one of the subgroups or if the age adjusted p-value was >0.05. I am thinking that might be too much, but id like your thoughts

On Tue, Sep 23, 2014 at 9:49 AM, Quynh Tran <notifications@github.com mailto:notifications@github.com> wrote:

You should look at the normal table. It has all the comparisons. Yes, I can write that part.

— Reply to this email directly or view it on GitHub < https://github.com/BridgesLab/CushingAcromegalyStudy/issues/30#issuecomment-56531571>

.

— Reply to this email directly or view it on GitHub< https://github.com/BridgesLab/CushingAcromegalyStudy/issues/30#issuecomment-56533098>.

— Reply to this email directly or view it on GitHub https://github.com/BridgesLab/CushingAcromegalyStudy/issues/30#issuecomment-56538310 .