With the widespread applications of nanomaterials such as carbon nanotubes, there is a growing concern on the biosafety of these engineered nanoparticles, in particular their interactions with proteins. In molecular simulations of nanoparticle-protein interactions, the choice of empirical parameters (force fields) plays a decisive role, and thus is of great importance and should be examined carefully before wider applications. Here we compare three commonly used force fields, CHARMM, OPLSAA, and AMBER in study of the competitive binding of a single wall carbon nanotube (SWCNT) with a native proline-rich motif (PRM) ligand on its target protein SH3 domain, a ubiquitous protein-protein interaction mediator involved in signaling and regulatory pathways. We find that the SWCNT displays a general preference over the PRM in binding with SH3 domain in all the three force fields examined, although the degree of preference can be somewhat different, with the AMBER force field showing the highest preference. The SWCNT prevents the ligand from reaching its native binding pocket by (i) occupying the binding pocket directly, and (ii) binding with the ligand itself and then being trapped together onto some off-sites. The pi-pi stacking interactions between the SWCNT and aromatic residues are found to play a significant role in its binding to the SH3 domain in all the three force fields. Further analyses show that even the SWCNT-ligand binding can also be relatively more stable than the native ligand-protein binding, indicating a serious potential disruption to the protein SH3 function.