Association of transmembrane (TM) helices taking place in the cell membrane has an important contribution to the biological function of bitopic proteins, among which receptor tyrosine kinases represent a typical example and a potent target for medical applications. Since this process depends on a complex interplay of different factors (primary structures of TM domains and juxtamembrane regions, composition and phase of the local membrane environment, etc.), it is still far from being fully understood. Here, we present a computational modeling framework, which we have applied to systematically analyze dimerization of 18 TM helical homo- and heterodimers of different bitopic proteins, including the family of epidermal growth factor receptors (ErbBs). For this purpose, we have developed a novel surface-based modeling approach, which not only is able to predict particular conformations of TM dimers in good agreement with experiment but also provides screening of their conformational heterogeneity. Using all-atom molecular dynamics simulations of several of the predicted dimers in different model membranes, we have elucidated a putative role of the environment in selection of particular conformations. Simulation results clearly show that each particular bilayer preferentially stabilizes one of possible dimer conformations, and that the energy gain depends on the interplay between structural properties of the protein and the membrane. Moreover, the character of protein-driven perturbations of the bilayer is reflected in the contribution of a particular membrane to the free energy gain. We have found that the approximated dimerization strength for ErbBs family can be related to their oncogenic ability.