Modeling membrane protein (MP) folding, insertion, association and their interactions with other proteins, lipids, and drugs requires accurate transfer free energies (TFEs). Various TFE scales have been derived to quantify the energy required or released to insert an amino acid or protein into the membrane. Experimental measurement of TFEs is challenging, and only few scales were extended to depth-dependent energetic profiles. Statistical approaches can be used to derive such potentials; however, this requires a sufficient number of MP structures. Furthermore, MPs are tightly coupled to bilayers that are heterogeneous in terms of lipid composition, asymmetry, and protein content between organisms and organelles. Here we derived asymmetric implicit membrane potentials from β-barrel and α-helical MPs and use them to predict topology, depth and orientation of proteins in the membrane. Our data confirm the 'charge-outside' and 'positive-inside' rules for β-barrels and α-helical proteins, respectively. We find that the β-barrel profiles have greater asymmetry than the ones from α-helical proteins, as a result of the different membrane architecture of gram-negative bacterial outer membranes and the existence of lipopolysaccharide in the outer leaflet. Our data further suggest that pore-facing residues in β-barrels have a larger contribution to membrane insertion and stability than previously suggested.