In the past few years, there has been an increased awareness about the role of diet in disease vulnerability. While individual studies have contributed fragments of information about the health effects of different foods, a comprehensive overview is missing. Systematic investigation of such empirical evidence from biomedical literature can facilitate valuable insights for diet-based disease intervention and drug discovery. In DietRx, we have assimilated the beneficial and adverse effects of 1781 foods on various diseases via text-mining from more than 27 million MEDLINE abstracts. Our relation extraction pipeline based on state-of- the-art named entity recognition tools and a highly accurate deep learning-based relation classifier significantly improvises upon the work of previous studies. We further linked data of 6992 food-chemicals and their disease/gene associations to investigate the underlying mechanisms of food-disease associations. Through information of the genes and foods connected with a disease, we formulate hypotheses of food-gene associations and make them available for investigation by the biomedical research community.