We have developed a new molecular dynamics (MD) based method for describing analytically local potentials at mobile N-H sites in proteins. Here we apply it to the monomer and dimer of the Rho GTPase binding domain (RBD) of the transmembrane receptor plexin-B1 to gain insight into dimerization, which can compete with Rho GTPase binding. In our method, the local potential is given by linear combinations, u((DL,K)), of the real combinations of the Wigner rotation matrix elements, D-L,D-K, with L = 1-4 and appropriate symmetry. The combination that "fits best" the corresponding MD potential of mean force, u((MD)), is the potential we are seeking, u((DL,K) (- BEST)). For practical reasons the fitting process involves probability distributions, P-eq proportional to exp(-u), instead of potentials, u. The symmetry of the potential, u((L,K)), may be related to the irreducible representations of the D-2h point group. The monomer (dimer) potentials have mostly A(g) and B-2u (B-1u and B-2u) symmetry. For the monomer, the associated probability distributions are generally dispersed in space, shallow, and centered at the "reference N-H orientation" (defined in section 3.1. below); for the dimer many are more concentrated, deep and centered away from the "reference N-H orientation". The u((DL,K)) functions provide a consistent description of the potential energy landscape at protein N-H sites. The L1-loop of the plexin-B1 RBD is not seen in the crystal structure, and many resonances of the L4 loop are missing in the NMR N-15-H-1 HSQC spectrum of the dimer; we suggest reasons for these features. An allosteric signal transmission pathway was reported previously for the monomer. We find that it has shallow N-H potentials at its ends, which become deeper as one proceeds toward the middle, complementing structurally the previously derived dynamic picture. Prospects of this study include correlating u((DL,K) (-) (BEST)) with MD force-fields, and using them without further adjustment in NMR relaxation analysis schemes.