The large amounts of proteomic data now available for cancer can used to investigate whether the physicochemical conditions of tumors are reflected in patterns of protein expression and chemical composition. Compositional analysis of more than 250 datasets for differentially expressed proteins compiled from the literature reveals a clear signal of higher stoichiometric hydration state (n H 2 O , derived from the theoretical formation reactions of proteins from particular basis species) in specific cancer types compared to normal tissue; this trend is also evident in pan-cancer transcriptomic and proteomic datasets from The Cancer Genome Atlas and Human Protein Atlas. In marked contrast to cancer, n H 2 O decreases for differentially expressed proteins in hyperosmotic stress (including high glucose) experiments and 3D cell culture compared to monolayer growth. Compositional analysis combined with gene ages (phylostrata) taken from the literature shows higher n H 2 O of human proteins earlier in evolution. Further analyses using amino acid biosynthetic reactions supports the conclusion that a net increase of water going into the reactions of protein synthesis is a biochemical characteristic shared by most cancer types. These findings raise the possibility of a basic physicochemical link between increased water content in tumors and the atavistic or embryonic patterns of gene and protein expression in cancer.