Efficient interatomic descriptors for accurate machine learning force fields of extended molecules.

Machine learning force fields (MLFFs) are gradually evolving towards enabling molecular dynamics simulations of molecules and materials with ab initio accuracy but at a small fraction of the computational cost. However, several challenges remain to be addressed to enable predictive MLFF simulations of realistic molecules, including: (1) developing efficient descriptors for non-local interatomic interactions, which are essential to capture long-range molecular fluctuations, and (2) reducing the dimensionality of the descriptors to enhance the applicability and interpretability of MLFFs. Here we propose an automatized approach to substantially reduce the number of interatomic descriptor features while preserving the accuracy and increasing the efficiency of MLFFs. To simultaneously address the two stated challenges, we illustrate our approach on the example of the global GDML MLFF. We found that non-local features (atoms separated by as far as 15 Å in studied systems) are crucial to retain the overall accuracy of the MLFF for peptides, DNA base pairs, fatty acids, and supramolecular complexes. Interestingly, the number of required non-local features in the reduced descriptors becomes comparable to the number of local interatomic features (those below 5 Å). These results pave the way to constructing global molecular MLFFs whose cost increases linearly, instead of quadratically, with system size.

Accurate global machine learning force fields for molecules with hundreds of atoms.

Global machine learning force fields, with the capacity to capture collective interactions in molecular systems, now scale up to a few dozen atoms due to considerable growth of model complexity with system size. For larger molecules, locality assumptions are introduced, with the consequence that nonlocal interactions are not described. Here, we develop an exact iterative approach to train global symmetric gradient domain machine learning (sGDML) force fields (FFs) for several hundred atoms, without resorting to any potentially uncontrolled approximations. All atomic degrees of freedom remain correlated in the global sGDML FF, allowing the accurate description of complex molecules and materials that present phenomena with far-reaching characteristic correlation lengths. We assess the accuracy and efficiency of sGDML on a newly developed MD22 benchmark dataset containing molecules from 42 to 370 atoms. The robustness of our approach is demonstrated in nanosecond path-integral molecular dynamics simulations for supramolecular complexes in the MD22 dataset.

Light Driven Ultrafast Bioinspired Molecular Motors: Steering and Accelerating Photoisomerization Dynamics of Retinal.

Photoisomerization of retinal protonated Schiff base in microbial and animal rhodopsins are strikingly ultrafast and highly specific. Both protein environments provide conditions for fine-tuning the photochemistry of their chromophores. Here, by combining time-resolved action absorption spectroscopy and high-level electronic structure theory, we show that similar control can be gained in a synthetically engineered retinal chromophore. By locking the dimethylated retinal Schiff base at the C11═C12 double bond in its trans configuration (L-RSB), the excited-state decay is rendered from a slow picosecond to an ultrafast subpicosecond regime in the gas phase. Steric hindrance and pretwisting of L-RSB are found to be important for a significant reduction in the excited-state energy barriers, where isomerization of the locked chromophore proceeds along C9═C10 rather than the preferred C11═C12 isomerization path. Remarkably, the accelerated excited-state dynamics also becomes steered. We show that L-RSB is capable of unidirectional 360° rotation from all-trans to 9-cis and from 9-cis to all-trans in only two distinct steps induced by consecutive absorption of two 600 nm photons. This opens a way for the rational design of red-light-driven ultrafast molecular rotary motors based on locked retinal chromophores.

Controlling Light-Induced Proton Transfer from the GFP Chromophore.

The front cover artwork is provided by the groups of Assoc. Prof. Anastasia V. Bochenkova (Lomonosov Moscow State University) and Prof. Lars H. Andersen (Aarhus University). The image shows the quantum nature of wavelength-dependent excited-state proton transfer in gas-phase H-bonded complexes of the GFP chromophore with an anionic proton acceptor. Read the full text of the Article at 10.1002/cphc.202100068.

Controlling Light-Induced Proton Transfer from the GFP Chromophore.

Green Fluorescent Protein (GFP) is known to undergo excited-state proton transfer (ESPT). Formation of a short H-bond favors ultrafast ESPT in GFP-like proteins, such as the GFP S65T/H148D mutant, but the detailed mechanism and its quantum nature remain to be resolved. Here we study in vacuo, light-induced proton transfer from the GFP chromophore in hydrogen-bonded complexes with two anionic proton acceptors, I- and deprotonated trichloroacetic acid (TCA- ). We address the role of the strong H-bond and the quantum mechanical proton-density distribution in the excited state, which determines the proton-transfer probability. Our study shows that chemical modifications to the molecular network drastically change the proton-transfer probability and it can become strongly wavelength dependent. The proton-transfer branching ratio is found to be 60 % for the TCA complex and 10 % for the iodide complex, being highly dependent on the photon energy in the latter case. Using high-level ab initio calculations, we show that light-induced proton transfer takes place in S1 , revealing intrinsic photoacid properties of the isolated GFP chromophore in strongly bound H-bonded complexes. ESPT is found to be very sensitive to the topography of the highly anharmonic potential in S1 , depending on the quantum-density distribution upon vibrational excitation. We also show that the S1 potential-energy surface, and hence excited-state proton transfer, can be controlled by altering the chromophore microenvironment.

A General Mechanism of Green-to-Red Photoconversions of GFP.

Here we dissect the phenomena of oxidative and reductive green-to-red photoconversion of the Green Fluorescent Protein. We characterize distinct orange- and red-emitting forms (λabs/λem = 490/565 nm; λabs/λem = 535/600 nm) arising during the Enhanced Green Fluorescent Protein (EGFP) photoconversion under low-oxygen conditions in the presence of reductants. These forms spectroscopically differ from that observed previously in oxidative redding (λabs/λem = 575/607 nm). We also report on a new green-emitting state (λabs/λem = 405/525 nm), which is formed upon photoconversion under the low-oxygen conditions. Based on the spectral properties of these forms, their light-independent time evolution, and the high-level computational studies, we provide a structural basis for various photoproducts. Under the low-oxygen conditions, the neutral quinoid-like structure formed via a two-electron oxidation process is found to be a key intermediate and a most likely candidate for the novel green-emitting state of the chromophore. The observed large Stokes shift is traced to the formation of the zwitterionic form of the chromophore in the excited state. Subsequently, this form undergoes two types of cyclization reactions, resulting in the formation of either the orange-emitting state (λabs/λem = 490/565 nm) or the red-emitting form (λabs/λem = 535/600 nm). The T65G mutant lacks one of the proposed cyclization pathways and, indeed, the photoconverted T65G EGFP exhibits a single orange-emitting state. In oxidative redding, the red-emitting state resembles the structure of the chromophore from asFP595 (λabs/λem = 572/595 nm), which is directly formed upon two-electron oxidation and deprotonation bypassing the formation of the quinoid-like structure. Our results disclose a general "oxidative" mechanism of various green-to-red photoconversions of EGFP, providing a link between oxidative redding and the photoconversion under low-oxygen conditions.