I dont think there should be a specific molecular marker. However, a battery of various molecular moieties may serve as a guiding light. To this effect Bhaskar has given you a fairly good list. I would suggest to evaluate their expression. The big problem, however, is that the test can be carried in blood which may not be the actual representative surrogate of the ocular tissues.
I wonder why do you need a molecular marker for diabetic retinopathy. It is just a fundus examination and you make a rough estimate quickly. May be an OCT may help further. But to look for a molecular marker in this case will have certain difficulties as blood may not be the true surrogate for retina.
Regarding nephropathy, I think the question becomes important. I suggest to do a pilot study on NMR of urine samples and correlate the metabolic profile with other nephropathies and healthy controls. A sample size of around 30 in each group should give a valid estimate of whether we can have any molecular, non-invasive marker of diabetic nephropathy.
Hope this is useful. If you have any specific question, I would be happy to discuss.