P-glycoprotein (Pgp) is an ATP-binding cassette transporter that eliminates toxins from the cell but causes multidrug resistance in chemotherapies. The crystal structures of Pgp revealed drug-like compounds bound to an inward-facing conformation in which the energy-harnessing nucleotide binding domains (NBDs) were widely separated with no interfacial interaction. Following drug binding, inward-facing Pgp must transition to an NBD dimer conformation to achieve ATP binding and hydrolysis at canonical sites defined by both halves of the interface. However, given the high degree of flexibility shown for this transporter, it is difficult to envision how NBDs overcome entropic considerations for achieving proper alignment in order to form the canonical ATP binding site. We explored the hypothesis that substrate occupancy of the polyspecific drug-binding cavity plays a role in the proper alignment of NBDs using computational approaches. We conducted twelve atomistic molecular dynamics (MD) simulations (100–300 ns) on inward-facing Pgp in a lipid bilayer with and without small molecule substrates to ascertain effects of drug occupancy on NBD dimerization. Both apo- and drug-occupied simulations showed NBDs approaching each other compared to the crystal structures. Apo-Pgp reached a pseudo-dimerization in which NBD signature motifs for ATP binding exhibited a significant misalignment during closure. In contrast, occupancy of three established substrates positioned by molecular docking achieved NBD alignment that was much more compatible with a canonical NBD dimerization trajectory. Additionally, aromatic amino acids, known to confer the polyspecific drug-binding characteristic of the internal pocket, may also govern polyspecific drug access to the cavity. The enrichment of aromatics comprising the TM4-TM6 portal suggested a preferential pathway over the aromatic-poor TM10-TM12 for lateral drug entry from the lipid bilayer. Our study also suggested that drug polyspecificity is enhanced due to a synergism between multiple drug-domain interactions involving 36 residues identified in TM1, 5, 6, 7, 11 and 12.