I am very new to processing and analyzing microarray data so forgive me if my question isn't well formulated. I have pre-processed and normalized my microarray data (Affymetrix U133a) so now I have 1 file containing normalized values with gene names in the rows and samples in the columns. The data consists of 8 subjects with expression taken at two time points (before and after sleep deprivation). I also have covariate data consisting of dichotomized results (0/1 pass/fail) from various neuropsychological tests that were administrated before and after sleep deprivation.
I am trying to find genes/gene clusters that are associated with sleep deprivation. I have looked numerous times at the limma userguide for guidance on how to set up my analysis, but haven't been able to figure this out.
Any guidance would be very appreciated. Thank you.
You have a relatively complex experiment, and you could ask questions more sophisticated than 'what genes are affected by sleep deprivation'. For example, you could say that the people who failed the neuro tests might be different in some way from those who passed, and you could then look for genes that are affected differently by sleep deprivation in that group (or alternatively, those genes that aren't affected by sleep deprivation in the people who passed the neuro exam).
And if there are genes that react differently to sleep deprivation in your pass/fail groups, then naively looking for genes that are affected by sleep deprivation may not pick those genes up. The easiest way to parameterize this experiment would be to estimate covariates for each group and then make contrasts amongst those groups. The analogous analysis in the limma User's Guide would be section 9.5, particularly 9.5.2.
If this is confusing to you, then I would recommend finding somebody local who can help you. Anybody can go to the tool store and get a set of tools to maintain their car, but pretty much nobody does that, because cars are complex and there are mechanics you can pay to do it for you. The same thing is true of data analysis. It's not so simple, and it's pretty easy to make mistakes if you don't know what you are doing. The smart play is often to find somebody who does know what's up and have them help you.
Thank you James! I really appreciate your response. The examples you mentioned are the exact questions I am trying to ask - I want to know which genes react differently to sleep deprivation and how that correlates to the neuropsychological test outcomes. Since posting my question I have realized that I will need to make contrasts among my experimental groups, as you also mentioned. Now I am trying to figure out my design matrix.
I completely agree with your tool store analogy. I foresee needing professional assistance soon, however with conservation of academic resources in mind I hope to completely understand the questions that need to be asked prior to infringing on my PI's budget. I will take a closer look at 9.5 and 9.5.2 in Limma.
Thank you James! I really appreciate your response. The examples you mentioned are the exact questions I am trying to ask - I want to know which genes react differently to sleep deprivation and how that correlates to the neuropsychological test outcomes. Since posting my question I have realized that I will need to make contrasts among my experimental groups, as you also mentioned. Now I am trying to figure out my design matrix.
I completely agree with your tool store analogy. I foresee needing professional assistance soon, however with conservation of academic resources in mind I hope to completely understand the questions that need to be asked prior to infringing on my PI's budget. I will take a closer look at 9.5 and 9.5.2 in Limma.
Thanks again!