Molecular Geometry Impact on Deep Learning Predictions of Inverted Singlet-Triplet Gaps.

We present a deep learning model able to predict excited singlet-triplet gaps with a mean absolute error (MAE) of ≈20 meV to obtain potential inverted singlet-triplet (IST) candidates. We exploit cutting-edge spherical message passing graph neural networks designed specifically for generating 3D graph representations in molecular learning. In a nutshell, the model takes as input a list of unsaturated heavy atom Cartesian coordinates and atomic numbers, producing singlet-triplet gaps as output. We exploited available large data collections to train the model on ≈40,000 heterogeneous density functional theory (DFT) geometries with available ADC(2)/cc-pVDZ singlet-triplet gaps. We ascertain the predictive power of the model from a quantitative perspective obtaining predictions on a test set of ≈14,000 molecules, whose geometries have been generated at DFT level (the same employed for the geometries in the training set), at GFN2-xTB level, and through Molecular Mechanics. We notice performance degradation upon switching to lower-quality geometries, with GFN2-xTB ones maintaining satisfactory results (MAE ≈ 50 meV on GFN2-xTB geometries, MAE ≈ 180 meV on generalized AMBER force field geometries), hinting at caution when dealing with specific chemical classes. Finally, we verify the performance of the model from the qualitative point of view, obtaining predictions on a different data set of ≈15,000 molecules already used to identify new IST molecules. We obtained predictions using both DFT and experimental X-ray geometries, with results on IST candidates similar to those provided by quantum chemical methods, with clear hints for the path toward improved performance.

Picosecond quantum-classical dynamics reveals that the coexistence of light-induced microbial and animal chromophore rotary motion modulates the isomerization quantum yield of heliorhodopsin.

Rhodopsins are light-responsive proteins forming two vast and evolutionary distinct superfamilies whose functions are invariably triggered by the photoisomerization of a single retinal chromophore. In 2018 a third widespread superfamily of rhodopsins called heliorhodopsins was discovered using functional metagenomics. Heliorhodopsins, with their markedly different structural features with respect to the animal and microbial superfamilies, offer an opportunity to study how evolution has manipulated the chromophore photoisomerization to achieve adaptation. One question is related to the mechanism of such a reaction and how it differs from that of animal and microbial rhodopsins. To address this question, we use hundreds of quantum-classical trajectories to simulate the spectroscopically documented picosecond light-induced dynamics of a heliorhodopsin from the archaea thermoplasmatales archaeon (TaHeR). We show that, consistently with the observations, the trajectories reveal two excited state decay channels. However, inconsistently with previous hypotheses, only one channel is associated with the -C13C14- rotation of microbial rhodopsins while the second channel is characterized by the -C11C12- rotation typical of animal rhodopsins. The fact that such -C11C12- rotation is aborted upon decay and ground state relaxation, explains why illumination of TaHeR only produces the 13-cis isomer with a low quantum efficiency. We argue that the documented lack of regioselectivity in double-bond excited state twisting motion is the result of an "adaptation" that could be completely lost via specific residue substitutions modulating the steric hindrance experienced along the isomerization motion.

Comparative Study of Uracil Excited-State Photophysics in Water and Acetonitrile via RMS-CASPT2-Driven Quantum-Classical Trajectories.

We present a nonadiabatic molecular dynamics study of the ultrafast processes occurring in uracil upon UV light absorption, leading to electronic excitation and subsequent nonradiative decay. Previous studies have indicated that the mechanistic details of this process are drastically different depending on whether the process takes place in the gas phase, acetonitrile, or water. However, such results have been produced using quantum chemical methods that did not incorporate both static and dynamic electron correlation. In order to assess the previously proposed mechanisms, we simulate the photodynamics of uracil in the three environments mentioned above using quantum-classical trajectories and, for solvated uracil, hybrid quantum mechanics/molecular mechanics (QM/MM) models driven by the rotated multistate complete active space second-order perturbation (RMS-CASPT2) method. To do so, we exploit the gradient recently made available in OpenMolcas and compare the results to those obtained using the complete active space self-consistent field (CASSCF) method only accounting for static electron correlation. We show that RMS-CASPT2 produces, in general, a mechanistic picture different from the one obtained at the CASSCF level but confirms the hypothesis advanced on the basis of previous ROKS and TDDFT studies thus highlighting the importance of incorporating dynamic electron correlation in the investigation of ultrafast electronic deactivation processes.

Assessment of the Electron Correlation Treatment on the Quantum-Classical Dynamics of Retinal Protonated Schiff Base Models: XMS-CASPT2, RMS-CASPT2, and REKS Methods.

We compare the performance of three different multiconfigurational wave function-based electronic structure methods and two implementations of the spin-restricted ensemble-referenced Kohn-Sham (REKS) method. The study is characterized by three features: (i) it uses a small set of quantum-classical trajectories rather than potential energy surface mapping, (ii) it focuses, exclusively, on the photoisomerization of retinal protonated Schiff base models, and (iii) it probes the effect of both methyl substitution and the increase in length of the conjugate π-system. For each tested method, the corresponding analytical gradients are used to drive the quantum-classical (Tully's FSSH method) trajectory propagation, including the recent multistate XMS-CASPT2 and RMS-CASPT2 gradients. It is shown that while CASSCF, XMS-CASPT2, and RMS-CASPT2 yield consistent photoisomerization dynamics descriptions, REKS produces, in some of these systems, qualitatively different behavior that is attributed to a flatter and topographically different excited state potential energy surface. The origin of this behavior can be traced back to the effect of the employed density functional approximation. The above studies are further expanded by benchmarking, at the CASSCF and REKS levels, the electronic structure methods using a QM/MM model of the visual pigment rhodopsin.

Automated QM/MM Screening of Rhodopsin Variants with Enhanced Fluorescence.

We present a computational protocol for the fast and automated screening of excited-state hybrid quantum mechanics/molecular mechanics (QM/MM) models of rhodopsins to be used as fluorescent probes based on the automatic rhodopsin modeling protocol (a-ARM). Such "a-ARM fluorescence screening protocol" is implemented through a general Python-based driver, PyARM, that is also proposed here. The implementation and performance of the protocol are benchmarked using different sets of rhodopsin variants whose absorption and, more relevantly, emission spectra have been experimentally measured. We show that, despite important limitations that make unsafe to use it as a black-box tool, the protocol reproduces the observed trends in fluorescence and it is capable of selecting novel potentially fluorescent rhodopsins. We also show that the protocol can be used in mechanistic investigations to discern fluorescence enhancement effects associated with a near degeneracy of the S1/S2 states or, alternatively, with a barrier generated via coupling of the S0/S1 wave functions.

Retinal chromophore charge delocalization and confinement explain the extreme photophysics of Neorhodopsin.

The understanding of how the rhodopsin sequence can be modified to exactly modulate the spectroscopic properties of its retinal chromophore, is a prerequisite for the rational design of more effective optogenetic tools. One key problem is that of establishing the rules to be satisfied for achieving highly fluorescent rhodopsins with a near infrared absorption. In the present paper we use multi-configurational quantum chemistry to construct a computer model of a recently discovered natural rhodopsin, Neorhodopsin, displaying exactly such properties. We show that the model, that successfully replicates the relevant experimental observables, unveils a geometrical and electronic structure of the chromophore featuring a highly diffuse charge distribution along its conjugated chain. The same model reveals that a charge confinement process occurring along the chromophore excited state isomerization coordinate, is the primary cause of the observed fluorescence enhancement.

On the fluorescence enhancement of arch neuronal optogenetic reporters.

The lack of a theory capable of connecting the amino acid sequence of a light-absorbing protein with its fluorescence brightness is hampering the development of tools for understanding neuronal communications. Here we demonstrate that a theory can be established by constructing quantum chemical models of a set of Archaerhodopsin reporters in their electronically excited state. We found that the experimentally observed increase in fluorescence quantum yield is proportional to the computed decrease in energy difference between the fluorescent state and a nearby photoisomerization channel leading to an exotic diradical of the protein chromophore. This finding will ultimately support the development of technologies for searching novel fluorescent rhodopsin variants and unveil electrostatic changes that make light emission brighter and brighter.

Evolution of the Automatic Rhodopsin Modeling (ARM) Protocol.

In recent years, photoactive proteins such as rhodopsins have become a common target for cutting-edge research in the field of optogenetics. Alongside wet-lab research, computational methods are also developing rapidly to provide the necessary tools to analyze and rationalize experimental results and, most of all, drive the design of novel systems. The Automatic Rhodopsin Modeling (ARM) protocol is focused on providing exactly the necessary computational tools to study rhodopsins, those being either natural or resulting from mutations. The code has evolved along the years to finally provide results that are reproducible by any user, accurate and reliable so as to replicate experimental trends. Furthermore, the code is efficient in terms of necessary computing resources and time, and scalable in terms of both number of concurrent calculations as well as features. In this review, we will show how the code underlying ARM achieved each of these properties.

Pro219 is an electrostatic color determinant in the light-driven sodium pump KR2.

Color tuning in animal and microbial rhodopsins has attracted the interest of many researchers, as the color of their common retinal chromophores is modulated by the amino acid residues forming the chromophore cavity. Critical cavity amino acid residues are often called "color switches", as the rhodopsin color is effectively tuned through their substitution. Well-known color switches are the L/Q and A/TS switches located in the C and G helices of the microbial rhodopsin structure respectively. Recently, we reported on a third G/P switch located in the F helix of the light-driven sodium pumps of KR2 and JsNaR causing substantial spectral red-shifts in the latter with respect to the former. In order to investigate the molecular-level mechanism driving such switching function, here we present an exhaustive mutation, spectroscopic and computational investigation of the P219X mutant set of KR2. To do so, we study the changes in the absorption band of the 19 possible mutants and construct, semi-automatically, the corresponding hybrid quantum mechanics/molecular mechanics models. We found that the P219X feature a red-shifted light absorption with the only exception of P219R. The analysis of the corresponding models indicate that the G/P switch induces red-shifting variations via electrostatic interactions, while replacement-induced chromophore geometrical (steric) distortions play a minor role. However, the same analysis indicates that the P219R blue-shifted variant has a more complex origin involving both electrostatic and steric changes accompanied by protonation state and hydrogen bond networks modifications. These results make it difficult to extract simple rules or formulate theories for predicting how a switch operates without considering the atomistic details and environmental consequences of the side chain replacement.