Fig. 4: TPS of water nanopore nucleation in membranes of Charmm36 DMPC lipids.

a Efficiency of AIMMD measured by the difference of expected and generated TPs, \(\Delta n={n}_{\exp }-{n}_{{\mbox{gen}}}\), and the ratio between simulation time of new transitions, \({T}_{{\mbox{TP}}},\) and total simulation time, \({T}_{{{\rm{all}}}}\). b Accuracy of committor models, correlating pore reaction coordinate \({\xi }_{{\mbox{P}}}\) to closest lipid displacement \({z}_{1}\) and the depletion of P and N atoms \({\Delta z}_{{{\rm{NP}}}1-4}^{\max }\). Boxes show median and \(25/75\)^th percentile of 800 bootstraps drawn from a total of 1000 MC steps; whiskers show \(2.5/97.5\)^th percentile. c Pore nucleation mechanism. The center plot shows the TPE for nucleation of a pore (state \({{\mathscr{P}}}\)) starting from a flat membrane (state \({{\mathscr{F}}}\)), projected onto the plane spanned by \({\xi }_{{\mbox{P}}}\) and \({z}_{1}\). Side panels show representative structures.