Deer antler has been widely used as a dietary supplement for hundreds of years in Asian countries. The chemical composition of deer antlers strongly depends on the growth conditions of the deer, especially the feeds, but the effects of different feeds on deer antlers have not been studied. To expand our knowledge of the chemical constituents of deer antler and establish an efficient way of differentiating antlers obtained with different feeds, we applied an NMR-based metabolomics approach and OPLS-DA multivariate analysis. We show that the antlers from one species on two different feeds, made from grass or mulberry trees, can be reliably differentiated by our metabolomics approach. We identified chemical constituents of the deer antlers and the marker compounds that contribute to the difference between the feed groups. We also rigorously validated our differentiation approach by showing that it can correctly classify blind samples into their respective feed groups. Our approach is expected to help design feeds to produce antlers with more defined constituents, especially those with higher bioactivities.