 Cheshire is a novel deep learning-based approach for predicting missing reactions in genome scale metabolic models, GEMs, based on their topological structure alone. It outperformed existing topology based methods in predicting artificially removed reactions over 926 high and intermediate quality GEMs. Furthermore, it was also able to improve the phenotypic predictions of 49 draft GEMs for fermentation products and amino acid secretion. This suggests that Cheshire is a powerful tool for GEM curation to reveal unknown links between reactions and observed metabolic phenotypes. This article was authored by Konchen, Chen Liao and Yang Yu Liu.