Is there any effective way to select generated antibody candidates (from a neural network), which were docked to antigen, using Rosetta Dock server?

Rosetta dock web server (https://r2.graylab.jhu.edu/auth/login?next=%2Fapps%2Fview%2Fdocking%2F8362) was used to select proper antibody candidates, and the thing is which exact quantitative features (except for Rosetta score and Lennard-Jones energy potential) I should consider while selecting those Ab structures, which would be able to bind to my Antigen target???