A model of the nicotinic acetylcholine receptor ion channel was elaborated based on the data from electron microscopy, affinity labeling, cysteine scanning, mutagenesis studies, and channel blockade. A restrained Monte Carlo minimization method was used for the calculations. Five identical M2 segments (the sequence EKMTLSISVL10LALTVFLLVI20V) were arranged in five-helix bundles with various geometrical profiles of the pore. For each bundle, energy profiles for chlorpromazine, QX-222, pentamethonium, and other blocking drugs pulled through the pore were calculated. An optimal model obtained allows all of the blockers free access to the pore, but retards them at the rings of residues known to contribute to the corresponding binding sites. In this model, M2 helices are necessarily kinked. They come into contact with each other at the cytoplasmic end but diverge at the synaptic end, where N-termini of M1 segments may contribute to the pore. The kinks disengage alpha-helical H-bonds between Ala12 and Ser8. The uncoupled lone electron pairs of Ser8 carbonyl oxygens protrude into the pore, forming a hydrophilic ring that may be important for the permeation of cations. A split network of H-bonds provides a flexibility to the chains Val9-Ala12, the numerous conformations of which form only two or three intrasegment H-bonds. The cross-ectional dimensions of the interface between the flexible chains vary essentially at the level of Leu11. We suggest that conformational transitions in the chains Val9-Ala12 are responsible for the channel gating, whereas rotations of more stable alpha-helical parts of M2 segments may be necessary to transfer the channel in the desensitized state.