In my molecular dynamics lecture, our prof said, that we always have to add hydrogen atoms to titratable groups, before we start the force field simulations, and that it is especially important for Histidine. But why is that? Why do we always have to add the hydrogen atoms? And do we only have to do that for titratable groups, or for all?
The RCSB Protein Data Bank (PDB) presently has 137,917 structures deposited in it, of which less than 10% have been solved by nuclear magnetic resonance (NMR). Most techniques in structural biology simply cannot resolve the positions of hydrogens in macromolecular structures; only NMR and neutron diffraction provide the positions of hydrogens "by default." Therefore, all-atom molecular dynamics force fields, the most notable of which are the OPLS, CHARMM, and AMBER series, require that the positions of hydrogens be deduced from the starting macromolecular structures and geometric criteria and patterns, that hydrogen bonds are known to generally follow, if the starting structure doesn't have hydrogens. "Generally" is an important word to point out here, as there often is no unique, unambiguous way to place the hydrogens in a structure. This is especially true of histidine, where the local environment of residues and/or cofactors may perturb its pKa in often unexpected ways. Simulating histidine in the wrong protonation state, when that histidine is an active site and crucial to the mechanism and process under study, may lead to spurious, misleading, and/or potentially nonsensical results. Hydrogens are required for a faithful, (relatively) realistic description of macromolecular structure, simply because they are always present in macromolecules and are crucial for macromolecular structure and functioning - recall that hydrogen bonds stabilize the secondary structure of proteins, the DNA double helix, etc., moreover, they often mediate intermacromolecular recognition and function. Thus, an explicit representation of hydrogens is required. The so-called coarse grained force fields, the most popular of which I believe are the various versions of MARTINI, group hydrogens along with heavier atoms to achieve an increase in the simulation time step and to reduce the number of particles in the system, but these force fields are a very poor choice when modeling specific recognition and interactions.
The ambiguities of protonation are not a matter of opinion, I'm afraid. One can go through all the structures in the PDB, only to find that the majority of them have no hydrogens at all. One can explore the link to the PDB I've provided; one will find that out of the 131,917 structures in the PDB, 123,461 have been resolved by X-ray crystallography, of which only 10,749 have a resolution below 1.5 Å. To resolve hydrogens, a resolution below 1 Å is required. Indeed, high resolution crystal structures, NMR structures, and neutron diffraction structures (only 60 in the entire PDB!) are less than 20% of the entire structural archive. Techniques like electron microscopy have resolutions between 4 and 10 Å; in such cases, people are unsure about the positions of the heavy atoms; they don't even dare think about the hydrogens. So, it's not just the titratable groups that are problematic, in most structures one can't see ANY hydrogens. Of course, for backbones and alanine side chains, one probably has a pretty good guess where the hydrogens are, given the positions of the heavy atoms. But to answer the question - in MD, most often one needs to add hydrogens everywhere, not just in the titratable groups, because the starting PDB structure doesn't have any hydrogens. In fact, often times structures from the PDB have missing heavy backbone or side chain atoms, because those couldn't be resolved; sometimes termini are missing, sometimes large stretches of protein; these often need to be modeled-in.