Logistic regression in metabolomics data So I have this huge dataframe with metabolomics data, meaning concentrations of 1030 metabolites, and other parameters like sex and age. Some of my data are from people with a particular disease (coded as 1), and control cases (0); and the final objective is to know if there is a correlation between any metabolites and the disease/no disease outcome (I am using the logistic regression for purely statistical reasons, not for classification).
I've been trying to get to understand how a logistic regression works, and how my data are, and how to make simple logistic regressions in R, which I did.  
My problems comes into play when I try to go from 4 variables to my 1000. I an utterly lost about how to do it. On top of that, I have to model my variables with age as confounding factor, that I know how it is but I haven´t found a place where they explain how to include this into my model.
Any ideas about how to proceed? 
 A: I'm going to assume that your 1000 explanatory variables are highly collinear, as is often the case with omics data (or so I hear).  This is to say that many of your variables are highly correlated with one another.  
When you have this situation, linear models (including logistic regression) have very unstable (variable) parameters.
One solution is regularization.  This will force your coefficients to shrink, and only the most important ones will remain.  L1 regularization can penalize coefficients down to zero.  
The downside is that your coefficient estimates are biased, and you won't be able to interpret the partial effects as such, like you would in an un-regularized model.  These methods are good for variable selection.
Another approach might be to fit a random forest, and then look at the variable importances.  These will tell you how much your predictive model suffers if a given variable is omitted, but they won't tell you what the marginal effect of that variable is on the outcome.  Furthermore, they don't assume, like L1-regularized linear models do, that the effect of the X's on y is a (transformed) linear function.
A textbook to get you started is here.
