The accurate experimental estimation of residence time (τ) of protein-ligand systems has become very relevant in drug design projects (1) due to its importance in the last stages of refinement of the drug's pharmacodynamics and pharmacokinetics (2). It is now well known that it is not sufficient to estimate the affinity of a protein-drug complex in the thermodynamic equilibrium process in in-vitro experiments (closed systems), where the concentrations of drug and protein remain constant. On the contrary, it is mandatory to consider the conformational dynamics of the system in terms of the binding and dissociation processes between protein and drugs in in vivo (open systems), where their concentrations are in constant flux. This last model has been proved to dictate much of several drugs' pharmacological activities in vivo (3). At the atomistic level, molecular dynamics simulations can explain why some drugs have better efficacy than others or the molecular aspects that make some drugs work better in one molecular target than others. Here, the protein kinases Aurora A/B, complexed with its inhibitor Danusertib, were studied using conventional and enhanced molecular dynamics (MD) simulations to estimate the dissociation paths and, therefore, the computational τ values and their comparison with experimental ones. These enzymes control different steps in the cell cycle, and their malfunction and dysregulation lead to the development of different types of cancer. Using classical molecular dynamics (cMD) and well-tempered metadynamics (WT-MetaD), tree differential residues within the Aurora A/B active site, which seems to play an essential role in the observed experimental Danusertib’s residence time against these kinases, were characterized. Using cMD, differences in the energetic contributions of residues THR217 and GLU161, which are located in the same structural position in Aurora A and Aurora B, respectively, to the binding of Danusertib were found. Moreover, a stable hydrogen bond between residues GLU161-LYS164 in Aurora B was identified, which provokes that LYS164 cannot move freely. In Aurora A, it was not observed a stable hydrogen bond between the analog residues THR217-ARG220, which caused that ARG220 can move with more flexibility. Then, using WT-MetaD and several replicas per system, the relative Danusertib‘s residence times against Aurora A/B kinases were measured on a nanosecond time scale, which was comparable to those τ values observed experimentally. In addition, the potential dissociation paths of Danusertib in Aurora A and B were characterized, finding differences that might be explained by the differential residues in the enzyme's active sites. In perspective, it is expected that this computational protocol can be applied to other protein-ligands complexes to understand at the molecular level the differences in residence times and amino acids that may contribute to it.