Engineering high chemo-, regio-, and stereoselectivity is a prerequisite for enzyme usage in organic synthesis. Cytochromes P450 can oxidize a broad range of substrates, including macrocycles, which are becoming popular scaffolds for therapeutic agents. However, a large conformational space explored by macrocycles not only reduces the selectivity of oxidation but also impairs computational enzyme design strategies based on docking and molecular dynamics (MD) simulations. We present a novel design workflow that uses enhanced-sampling Hamiltonian replica exchange (HREX) MD and focuses on quantifying the substrate binding for suggesting the mutations to be made. This computational approach is applied to P450 BM3 with the aim to shift regioselectively toward one of the numerous possible positions during β-cembrenediol oxidation. The predictions are experimentally tested and the resulting product distributions validate our design strategy, as single mutations led up to 5-fold regioselectivity increases. We thus conclude that the HREX-MD-based workflow is a promising tool for the identification of positions for mutagenesis aiming at P450 enzymes with improved regioselectivity.