Abstract
Electrostatic interactions are crucial for protein structure and function, especially in mesoscopic protein channels where ion selectivity is largely governed by the protein's electrostatic properties. Understanding the protonation state of ionizable residues across pH values -often described by their pKa- is key to linking structure and function. However, experimental pKa determination is challenging, typically carried out using Nuclear Magnetic Resonance only in a limited number of membrane proteins. Thus, computational methods are the primary alternative. Constant pH Molecular Dynamics (CpHMD) simulation is one of the most accurate pKa prediction methods in proteins that contain many charged residues since it captures the coupling between conformational dynamics and residue protonation. Here we study the charge state of a general diffusion porin, OmpF, in which protons exert a crucial regulation of the channel discrimination of small inorganic ions as well as antibiotic translocation. We compare different pKa prediction methods, using CpHMD as a benchmark, and discuss the somewhat unusual titration of several acidic residues. The most widely used pKa prediction methods, though effective for globular proteins, fall short for membrane-embedded channels either because they were trained using pKa measurements in globular proteins or because of a poor description of the lipidic environment.