Assessment of diet currently relies on self-reporting, such as food logs, 24 hour recalls and food frequency questionnaires. Self-reporting of diet is inaccurate due to memory lapses, lying and biased language. A molecular-based approach to assess diet would allow accurate reporting of diet for researchers, medical professionals and patients. We performed metaproteomic analysis of 22 human stool samples collected from four individuals enrolled at the Duke Diet and Fitness Center. Samples were collected between 01Aug19 and 22Nov19. All meals at the Duke Diet and Fitness Center were recorded such that the approximate quantities of each food item consumed are known.