Blueprinting, implementation, and application of fully automatic protocols for the QM/MM modeling of photo-excited states of rhodopsin variants

The semi-automatic building of QM/MM models of rhodopsins has been recently proposed, by means of a new technology called Automatic Rhodopsin Model protocol. In its original version, here called original ARM protocol, published in 2016, the constructed QM/MM models were found to be useful for biophysical studies and for protein engineering, but had the disadvantage of being time-consuming to prepare, error-prone, and, also, difficult to replicate when the same model was independently constructed by different investigators. These issues were a consequence of the <> (i.e., not fully automated) nature of the protocol, since (i) the generation of its input was achieved through <> of the template structure and (ii) the code (i.e., the computer program) for the construction of the QM/MM model was written as a non homogeneous collection of bash scripts, not driven by a parent program. Such methodological and computational pitfalls impaired the possibility of comparatively studying hundreds of rhodopsins (i.e., light-sensitive proteins belonging to the same superfamily), as well as hoping that, in the future, a similar protocol could be generalized to other families of light-responsive proteins (e.g., Xanthorhodopsin, phytochromes or synthetic rhodopsin mimics) of interest for biological or biotechnological applications. In order to overcome the above drawbacks, this doctoral Thesis is devoted to the design of a substantially improved ARM methodological framework, characterized by a fully automated, rather than manual, construction of QM/MM models. Accordingly, I introduce the blueprinting of four different ARM-based fully automatic protocols for the QM/MM modeling of rhodopsin electronically excited states. Furthermore, I present their implementation into a new, user-friendly, Python-based software package, called ARM software package, conceived for allowing the use of each protocol via a ``one-click'' command given either at the command-line or, in certain cases, Web-interface levels. Finally, I report on the performance of the four ARM-based protocols, highlighting both their methodological and scientific capabilities as well as their current limitations. To do so, I have constructed and employed a benchmark set of about 150 wild-type and mutant rhodopsins, as well as carried out selected applications, directed to the prediction of trends in light-induced properties, including absorption and emission spectra, as well as excited-state molecular dynamics. Such trends unveil different mechanistic aspects of color tuning and fluorescence emission, as well as, more in perspective, the systematic prediction of photoisomerization quantum yields. In conclusion, the research carried out during my doctoral Thesis has generated and explored novel, automated, ARM-based research tools and, most importantly, a programming framework called the ARM package. I believe that these tools and package have the potential to be generalized, thanks to their characteristics that I will thoroughly describe. In other words, my hope is that the research line started with my thesis will not only be useful for achieving better performing QM/MM models of rhodopsins but be expanded to deal with other sets of light-responsive proteins useful, for instance, in optogenetic studies.