Analyzing a clinical dataset through R

Hi everyone! I'd like to have some advice about how to do a regression analysis (on R) on a clinical dataset I'm working with. In this dataset, I have different kinds of patients: Some of them with the disease, other that risk to develop it, and still others from a control group. These are all in the same dataset and distinguished on a column by their code (some letters followed by the specific numeric id of the individual), that changes with respect to their group (control, disease, risk). Moreover, I've got a precise variable I have to work with, that's a rating scale of the disease. In addition to this, I'm required to use some biomarkers (and maybe biographical/familiar/social data, but especially and first of all biomarkers) obtained by some specific medical measurements done during the experiment, which are about 20 variables.
I'm required to do regression analysis evaluating also possible non-linearities, and to organize my data in a training and testing set.
Do you have some advice about how to work? Should I do a multiple regression through the biomarkers? What can be a good way to use them? Moreover, how can I use the fact that I have 3 groups (control,risk,disease)? Do you think I have to do distinguished analysis or something else?


Less is more. Stay pure. Stay poor.
Can you provide a small piece of your data, so we can see what you are working with? Also, how big is your sample size and how is "rating scale" formatted?
In this precise moment I'm not having the dataset available. However I can say that the sample is a bit more than a hundred of people, and that the rating scale is a discrete and ordinal one with a finite range of values. It aggregates only some items of the entire scale to detect the disease (in other words it's a part of the scale). This scale has not value on people of the control group (somebody said me it's normal because they need to be considered as asymptomatic and that for this I can assign to them the value 0 of the scale).