Previous studies have clustered and analyzed proximity sounding profiles associated with tornadic right-moving supercells. However, few have objectively determined the optimal number of clusters or identified key features driving cluster assignments. Using self-organizing maps, we group soundings based on virtual buoyancy and wind profiles. A novel recursive feature elimination reveals that virtual buoyancy at heights of approximately 0.5 km and 1.4 km (near the level of free convection) are the most significant features in this clustering, with both acting as a proxy for the potential strength of convective inhibition (CIN). The magnitude of virtual buoyancy at 4.7 km characterizes the potential maximum strength of buoyancy (convective available potential energy; CAPE). The virtual buoyancy magnitude at 9.3 km is also important, likely corresponding to the height of the equilibrium level (EL). Low-level (< 3 km) wind profiles, particularly the westerly component, also play a key role in determining cluster assignments.