Based on homology of partial sequences, on physico-chemical evidence and on studies using chemical modification, we came to the tentative conclusion that tetrameric glucose dehydrogenases from Bacillus megaterium and B. subtilis should have a structure closely related to that of lactate dehydrogenase. The overall homology of primary structures was found to be very low, however, and independent predictions of secondary structure produced a clearly different pattern of beta-strands and alpha-helices. We nevertheless tried a manual prediction based on the hydrophobic nature of internal beta-sheet and on the amphiphilic character of external helices. This treatment led to the identification of analogues of all the beta-strands present in lactate dehydrogenase with the exception of beta C. Six amphiphilic helices were identified corresponding to alpha B, alpha C, alpha D, alpha 1F, alpha 2F and alpha 3G in lactate dehydrogenase. Conserved functional residues were found at analogous positions. The Q and R intersubunit contacts could be identified and partial proteolysis was found to occur on the outer surface of the tetramer. The structure was found to explain the better binding of NADP as compared to NAD+ and offered a rationalization of the role of the essential lysine at position 201.