We have performed dynamic docking simulations between a flexible receptor and a highly flexible ligand by employing multicanonical molecular dynamics (McMD) [1-2]. We have applied our method to predict the native binding configuration and sample the intermediary binding structures between the enzyme β-secretase 1 with a wide binding pocket and its medium-sized inhibitor 3MR [3]. Representative structures located at free energy minima obtained from McMD were taken and subjected to canonical simulations to refine and validate them, reproducing the native complex structure in high agreement with the experimental data. In addition, the binding free energy was estimated by umbrella sampling (US) simulations along representative pathways obtained from the McMD ensemble, followed by weighted histogram analysis to estimate the affinity, which also reproduced the experimental inhibitory affinity. The sampled ensemble by the US simulations smoothly connected the bound and unbound states, refining the binding pathway while staying true to the McMD ensemble. Interestingly, the loss of interactions between the two molecules along the pathway were clearly shown in the free energy landscape, reiterating the fundamental importance of atomistic interactions to the binding affinity between receptor and drug. [1] N. Kamiya et al. Proteins 70, 41-53 (2008). [2] G.-J. Bekker et al. J. Chem. Theory Comput. 13, 2389-2399 (2017). [3] G.-J. Bekker et al. J. Phys. Chem. B 123, 2479-2490 (2019).