Good take as always. I don’t think this axis makes sense really though. I would argue a diffusion model is more physics based than a lennard jones forcefield. Harmonic approximation about the minima is in every physics text book but I’ve never seen a 1/r^12 repulsion.
Alex Zhavoronkov, PhD (aka Aleksandrs Zavoronkovs)
Learnt from experiment. vs learnt from quantum mechanical data is a sensible axis but it seems now that they’ve trained on the whole PDB that it will need every new source of training data possible to keep improving.
@TimothyDuignan
1/r^12 is a commonly used potential to model gas phase interactions. I presume it can be explained by perturbation theory of dipole interactions but I am not sure.
@YigithanGediz
Yes apparently Lenard jones just fit it to the noble gas phase data. Also used because it’s computationally cheap. Never seen a derivation
@CalvinMccarter
Yes good point but I would argue MD with LJ doesn’t tell you this either. LJ 1/r^6 parameters don’t agree with first principles c6 coefficients. So it can’t really tell you about the physical causes. It does give more physically plausible trajectories though that’s true.