models variability for AF multimer

Dear all,

I am thinking for a while about alphafold model variability for multimer models. Even if use 20 multimer predictions (100 models) the resulting complex prediction are very very similar, but in reality there should be be some variability in the models as the complex is accommodating its potential energy surface.

Is there a parameter in AF that would allow me to produce more "variability in the resutls of AF multimer?

Thanks a lot,
Fabian Glaser
Technion, Israel