On the application of Job's method of continuous variation to the stoichiometry of protein-ligand complexes.