Investigating the origin of Automatic Rhodopsin Modeling outliers using the microbial Gloeobacter rhodopsin as testbed

The Automatic Rhodopsin Modeling (ARM) approach is a computational workflow devised for the automatic build up of hybrid quantum mechanics/molecular mechanics (QM/MM) models of wild-type rhodopsins and mutants, with the purpose of establish- ing trends in their photophysical and photochemical properties. Despite the success of ARM for accurately describing the visible light absorption maxima of many rhodopsins, for few cases, called outliers, it might lead to large deviations with respect to exper- iments. Applying ARM to Gloeobacter Rhodopsin (GR), a microbial rhodopsin with important applications in optogenetics, we analyze the origin of such outliers in the absorption energies obtained for GR wild-type and mutants at neutral pH, with a total root mean square deviation (RMSD) of 0.42 eV with respect to the experimen- tal GR excitation energies. Having discussed the importance and the uncertainty of one particular amino-acid pKa , namely histidine at position 87, we propose and test several modifications to the standard ARM protocol: (i) improved pKa predictions along with the consideration of several protonation microstates, (ii) attenuation of the opsin electrostatic potential at short-range, (iii) substitution of the state-average com- plete active space (CAS) electronic structure method by its state-specific approach, and (iv) complete replacement of CAS with mixed-reference spin-flip time-dependent density functional theory (MRSF-TDDFT). The best RMSD result we obtain is 0.2 eV combining the protonation of H87 and using MRSF/CAMH-B3LYP.