I am trying to perform a differential gene expression of a big microarray data using Limma.
I found roast is the function of choice for when the sample size is large. however, i have some problem for selecting or tuning its parameters.
roast(y,index1,design,contrast=2)
Y is the matrix consisting of gene expressions (each row is a probe and each column a sample) design is my design of choice (in my case is a binary (controlled and untreated))
I could not find the way to define the index1
I could not find the way to define the contrast too is there any body able to help me out ?
The roast() function isn't designed for differential expression. Can you say why you might think that it is? Instead, roast() is designed for a particular type of gene set testing.
There is a comprehensive User's Guide for limma, that covers pretty much every imaginable scenario for differential testing. You should peruse that guide if you plan on using limma for analyzing data. If you find the User's Guide confusing, then I would highly recommend finding a local statistician to help you with your analyses.
This is the point at which you have to make a decision for yourself. You are struggling to figure out how to use limma (and make basic plots), which is the easy part. The more difficult part is to understand statistically what you are doing. If the code above is really what you have done so far, then you are also struggling with your understanding of the statistical part as well (which like I said, is the harder part of the process).
My recommendation that you find a local statistician still stands - that's the easy way to get the answers you want, and is probably the smart way to proceed. There is nothing wrong with paying someone for their expertise. I have a mechanic to fix my car, a doctor to care for my health, a lawyer to keep me out of jail, etc.
If however you want to proceed with the analysis, then you are going to have to shoulder much more of the burden. You need to figure out on your own how to make simple plots, and how to extract data from various things, and how to use limma to make the various comparisons. In addition, you have to ensure that you know enough statistics to keep out of trouble. The Dunning-Kruger effect can be a real problem in all fields, including statistics.
If you want to be responsible for the analysis of your data, then you have to become responsible for the analysis of your data. There is no other way.
@James W. MacDonald it is funny since I am a mathematician/statistician .
My problem is not how to plot a simple MA or volcano or another plot ( this is so easy to even ask because even if you don't understand statistic you can google and run it. Knowing R or doing statistics are two different things ! however, I would always recommend myself, if I don't know anything or if I don't want to do anything, I don't just comments on people question.
I simply needed to know whether there are other ways to interpret the results or not!
I believe such comment won't help anyone and I definitely do not agree with your comment
@James W. MacDonald you are right !
What i did so far, is as follows
Do you have any idea how i can make a plot or select those that are now statistically different ?
In limma tutorial they used MAplot, but if the data is huge such a plot is useless,
how would you export for example 10 most up-regulated and 10 most down-regulated genes ?
This is the point at which you have to make a decision for yourself. You are struggling to figure out how to use limma (and make basic plots), which is the easy part. The more difficult part is to understand statistically what you are doing. If the code above is really what you have done so far, then you are also struggling with your understanding of the statistical part as well (which like I said, is the harder part of the process).
My recommendation that you find a local statistician still stands - that's the easy way to get the answers you want, and is probably the smart way to proceed. There is nothing wrong with paying someone for their expertise. I have a mechanic to fix my car, a doctor to care for my health, a lawyer to keep me out of jail, etc.
If however you want to proceed with the analysis, then you are going to have to shoulder much more of the burden. You need to figure out on your own how to make simple plots, and how to extract data from various things, and how to use limma to make the various comparisons. In addition, you have to ensure that you know enough statistics to keep out of trouble. The Dunning-Kruger effect can be a real problem in all fields, including statistics.
If you want to be responsible for the analysis of your data, then you have to become responsible for the analysis of your data. There is no other way.
@James W. MacDonald it is funny since I am a mathematician/statistician .
My problem is not how to plot a simple MA or volcano or another plot ( this is so easy to even ask because even if you don't understand statistic you can google and run it. Knowing R or doing statistics are two different things ! however, I would always recommend myself, if I don't know anything or if I don't want to do anything, I don't just comments on people question.
I simply needed to know whether there are other ways to interpret the results or not!
I believe such comment won't help anyone and I definitely do not agree with your comment