I do not believe that the ddg monomer application is set up to do backrub sampling by itself, although it may be possible to set up a multi-application protocol which performs the backbone perturbation and then the monomer ddg in separate steps.
Note that protocol 16 in the Kellog paper does not use extensive backbone perturbation – the only backbone movement you’d see from that protocol would be movement from the constrained minimization.
Active perturbation of the backbone (e.g. with backrub) would be more akin to protocol 20, the one with Monte Carlo ensemble generation.