Early studies have shown that some of diterpenoid alkaloids, found in highly toxic plants of the genera Aconitum and Delphinium, act at neuronal nicotinic acetylcholine receptors (nAChRs) and exhibit potent N-cholinolytic activity. In the current study, GA-MLRA and GA-PLS approaches have been used to build QSAR models to predict N-cholinolytic activity measured in vivo (blockade of neuromuscular conductivity, BNMC and third eyelid relaxing activity, TYRA) and in vitro (suppression of frog's abdominal straight muscles on acetylcholine, SAM) for a series of diterpenoid alkaloids. Random splitting of a data set (five trials in total) produced QSAR models of a good level of correlation between experimental in vitro/in vivo and calculated N-cholinolytic activity expressed as log(1/ED(50)) with following average statistical parameters log BNMC (r(2) = 0.87, s = 0.14, q(2) = 0.82), log TYRA (r(2) = 0.80, s = 0.29, q(2) = 0.67), log SAM (r(2) = 0.84, s = 29, q(2) = 0.64). QSAR results suggest descriptors accounting for H-bond capability of molecules influence all three type of N-cholinolytic activity with additional contribution of steric and reactivity features as identified for TYRA and SAM data, respectively. The alkaloid-receptor complexes were further analyzed by means of AutoDock Vina docking program using the binding site of MLA complexed with AChBP (homolog of the ligand binding domain of nAChRs) as template. All compounds were shown to be well fitted in the binding pocket of native MLA with good correlation exhibited between their ED(50) and AutoDock Vina binding free energy. An analysis of the possible factors significant for the ligand recognition has been enhanced by comparative docking studies performed for structurally related lycoctonine-type alkaloids (lappaconitine and aconitine) that are known to bind to voltage-gated Na(+) channel, but not to nAChRs.