An atomic model of the sickle hemoglobin (HbS) fiber was synthesized by combining the molecular coordinates of the fiber (obtained from electron microscopy) with atomic coordinates of the sickle hemoglobin double strand (obtained from X-ray crystallography). The model is stereochemically acceptable. The majority of polymerization-sensitive HbS mutants are located at fiber contact sites and the majority of the mutants that do not affect polymerization are not located at contact sites. The residues at intermolecular contacts in the fiber model are reported. We have searched the coordinate space in the vicinity of the EM reconstructions to find models with alternative sets of coordinates that satisfy the mutant data, contain 5-Angstrom contacts between double strands, and are stereochemically acceptable. This involved a systematic examination over 297 different models. The alternative fiber models were generated with a range of fiber pitch, double-strand positions, and double-strand polarity. Models which had unacceptably close contacts between atoms, failed to satisfy the mutant data, or did not have 5-Angstrom contacts between double strands were considered unacceptable. None of the acceptable alternative fiber models improved the agreement between the polymerization behavior of HbS mutants and their contact site location. However, several models could account for the polymerization data equally well. Residue locations for single-site HbS mutations that could discriminate between alternative fiber models are proposed. The twist of HbS fibers varies in an apparent random manner with an average rotation of 7.8 +/- 2.5 degrees per molecule and a maximum rotation of 16 degrees per molecule. The number of inter-double-strand contacts as a function of fiber twist shows a broad maximum around 9 degrees and may account for the observed range of fiber pitch. This study shows that the upper limit on the fiber twist could result from a loss of axial contacts and repulsive van der Waals interactions between residues involved in interstrand contacts. The loss of axial contacts limits the radial growth of the fiber. In the appendix we analyze the methodology used by I. Cretegny and S. J. Edelstein [(1993) J. Mol. Biol. 230, 733-738] to build a model of the fiber. Our examination reveals shortcomings in the methodology of Cretegny and Edelstein. One result of these shortcomings is that the model synthesized by Cretegny and Edelstein is not stereochemically acceptable because it gives rise to a large number of excessively close (less than 1.4 Angstrom) atom-atom contacts, suggesting interpenetration of the molecular envelopes. (C) 1993 Academic Press, Inc.