Since the availability of experimentally determined structures, such as X-ray crystallography data from the Protein Data Bank (PDB), computational scientists have essentially been required to use these structures as initial models for simulations. The process of creating model structures varies widely among researchers, ranging from open source to commercial software, aimed at achieving physically and chemically accurate models. However, with multiple struc-tures in the PDB, there is no clear guideline for selecting the optimal one. This article summarizes, based on the author's experience, five key points to consider when choosing structures for simulations like molecular docking, molecular dynamics, and fragment molecular orbital (FMO) calculations.