Membrane pore formation by peptides is a multi-step event—initial peptide aggregation, embedding, and pore opening, followed by either pore expansion (cell lysis) or membrane fusion (viral or synaptic membrane fusion). Each step involves the coordinated movement of not just the peptides but also the water and lipids surrounding the peptides making it a long timescale event difficult to achieve with unbiased molecular dynamics (MD). Recent work from our lab showed that an asymmetric membrane creates lateral stress which in turn reduces the free energy (more favorable) of the pore opening and expanding, compared to a symmetric system. To compare this effect in different membrane systems, one would need to run hundreds of umbrella sampling simulations (US), each run for an aggregate of at least 6microsec. Recently, it has been shown that DDPMs are capable of recreating complete PMFs for small molecules while trained on a fraction of the trajectory data. We employ the DDPM to peptide-membrane system and compare its performance in generating peptide pore formation PMFs in symmetric/asymmetric membranes.
Bodosa et al. (Sun,) studied this question.