Hi, Joseph.

Minimizing energy before docking is not always necessary or convenient. This depends on many factors. Molecular Docking programs (grand majority) do not have capacity to move the protein backbone and if they do, have limitations. As you say, minimization of energy prior to molecular docking can change protein structure and therefore affect the outcome of your simulation. ideal is that protein is in conformation in which it would bind to that ligand in particular, but we do not always have this.

In case of cryo-em structures, although I do not have much experience in the matter, usually this technique was used with other techniques such as X-ray and NMR to solve macromolecular systems and had low resolution. In this way, each component is resolved separately with high resolution techniques and integrated into the system density map solved by cryo-em (generally with high symmetry, eg viruses). Currently, this technique already has capacity to solve structures to atomic resolutions (https://doi.org/10.1016/j.str.2018.04.004, among others).

Therefore, the analysis and treatment of protein structure depends on quality of model obtained and not on technique with which it is solved.

Docking simulations are a valuable approach, but have limitations. It is possible that you perform a docking simulation of a complex with a known structure and do not obtain a pose similar to experimental result, or at least, not the one that best predicts docking program. These limitations are known. Similar can happen when you get a conformation in a docking simulation and then in a simulation of molecular dynamics the ligand separates from protein.

Regarding if the structure that you get after minimizing energy is a reliable representation of the system, probably yes. Parts of a structure of a protein in complex with a ligand ?. It is known that proteins can change their structure once a ligand is bound; if the structure you get does not have problems such as defolding or things like that, this conformation is probably similar or is in transition to conformation of protein in aqueous solution.

I recommend you do a molecular docking simulation with structure without minimization and then a simulation of molecular dynamics of structure that you obtain. In this way you can analyze interaction between protein and ligand, for example: prevalence of hydrogen bonds, etc. If you need other elements, you can use meta-dynamics tools.

Best wishes

You have to keep in mind that proteins are flexible - any set of coordinates in the best case represents an average around which the actual coordinates fluctuate as a result of thermal motion. These fluctuation can be large and slow in the case of flexible loops. The actual conformation can be further influenced by crystal contacts and the far-from-physiological conditions conductive to crystallization.

In addition, many proteins show fairly large conformational changes between the liganded and the unliganded state.

Optimally, for docking you use not a single structure, but an ensemble of structures generated e.g. by molecular dynamics in the hope of capturing the one conformation best suited for complex formation. Particularly if your putative binding site is formed by loop and coil structures rather than heavily hydrogen-bonded beta-sheets and helices, if the local b-factors are high compared to the core of the protein or if the overall resolution is low, you should consider multiple conformations.

https://www.researchgate.net/post/Is_energy_minimization_is_necessary_prior_to_dock_a_molecule

https://www.researchgate.net/post/Is_energy_minimization_after_docking_necessary_before_running_analysis

https://www.researchgate.net/post/Why_do_we_perform_an_energy_minimization_step_for_protein

https://www.researchgate.net/post/Is_energy_minimization_of_a_ligand_and_a_protein_compulsory

https://www.researchgate.net/post/How_do_I_perform_energy_minimization_of_ligands_before_docking

In theory, there is absolutely no need to minimize ligand structures prior to feeding them to a docking engine. Every docking algorithm would anyway strongly distort ligands when searching for an optimal pose for them. However, certain software developers recommend to minimize. They usually refer to unpublished virtual screening studies that have shown better results with minimized ligands. I guess it's more of an urban legend than anything else. On the other hand, these days, everyone can afford this step prior to docking; takes only about a second per ligand with the modern hardware.

Awesome. Drs. Kireev, Khamis, Honegger and Mr. Tresanco, thank you for all the great answers. If I have any more I will be sure to post to this thread.