I’m currently performing MD analysis of receptor-cytokine interaction (ppi) within 200 nanoseconds. I’m testing single-point mutation on cytokines, and I want to capture differences in IMFs between the wt and mutant cytokine with the receptor ectodomain. The docking pose of the complex was generated on AF multimer, and disulfide bonds were added using pdb2gmx. I’m using the most recent CHARMM-36 forcefield, from february 2026.
I’ve been reading through some literature, but I cannot find an exact paper that fits my circumstances. How many nanoseconds do I need for valid quantification of IMF lifetimes, and other parameters in protein-protein interaction?
Additionally, if I want to perform binding/unbinding events MD in the future after sifting through single-point mutations, how many nanoseconds do I need for this MD simulations?
I don’t know what IMF means (intermolecular forces?), but the time required for a simulation depends heavily on the specific phenomenon/system you are interested in. For your particular cytokine-receptor system, I recommend checking the literature for measured dissociation rates (k_off) to have an estimate. A very rough ballpark estimate for protein-protein binding would be on the microsecond scale.
What I understand you are trying to do is to compare the binding affinity of the different cytokines. If that is really the case, I would strongly recommend following one of the many protocols for estimating binding ddG using any of the several biased MD protocols available (FEP, TI, Umbrella sampling, WHAM, MM-PBSA, metadynamics…). Even the coarsest of these methods (end-state methods like MM-PBSA) would likely provide a better estimate of the difference you’re trying to measure than just running unbiased MD. The amount of sampling you’d need to be able to tell the difference between these mutations with any statistical significance would be enormous.
As general guidelines, I highly recommend the paper “Avoiding False Positive Conclusions in Molecular Simulation” (https://pubs.acs.org/doi/10.1021/acs.jctc.8b00391) to understand how many simulations and how long they need to be for your particular case.