Skip to main content
NCBI home page
ACS Central Science logoLink to ACS Central Science
. 2024 Jan 31;10(2):302–314. doi: 10.1021/acscentsci.3c01403

Exploring Chemical Space Using Ab Initio Hyperreactor Dynamics

Alexandra Stan-Bernhardt †, Liubov Glinkina †, Andreas Hulm †, Christian Ochsenfeld †,‡,*
PMCID: PMC10906254  PMID: 38435517

Abstract

graphic file with name oc3c01403_0008.jpg

In recent years, first-principles exploration of chemical reaction space has provided valuable insights into intricate reaction networks. Here, we introduce ab initio hyperreactor dynamics, which enables rapid screening of the accessible chemical space from a given set of initial molecular species, predicting new synthetic routes that can potentially guide subsequent experimental studies. For this purpose, different hyperdynamics derived bias potentials are applied along with pressure-inducing spherical confinement of the molecular system in ab initio molecular dynamics simulations to efficiently enhance reactivity under mild conditions. To showcase the advantages and flexibility of the hyperreactor approach, we present a systematic study of the method’s parameters on a HCN toy model and apply it to a recently introduced experimental model for the prebiotic formation of glycinal and acetamide in interstellar ices, which yields results in line with experimental findings. In addition, we show how the developed framework enables the study of complicated transitions like the first step of a nonenzymatic DNA nucleoside synthesis in an aqueous environment, where the molecular fragmentation problem of earlier nanoreactor approaches is avoided.

Short abstract

We introduce hyperreactor dynamics to computationally navigate the vast chemical space of molecular systems on an elevated potential energy surface complemented by a pressure-inducing piston.

Introduction

While theoretical validation of experimental results has become indispensable, breakthroughs in chemistry are still often achieved by exhaustive experiments based on trial-and-error, which are both time- and resource-consuming. Lately, predictive approaches have been emerging as powerful acceleration tools in biology, (bio)chemistry, physics, and materials science, e.g., to drive drug discovery and catalyst design.1,2 In particular, complex fields with many competing hypotheses, like the origins of life under prebiotic conditions,3 might profit from theoretical prediction. How, when, and where did life originate and which prebiotic molecular species played the key roles to allow for the emergence of life’s building blocks: amino acids and peptides, ribose, (oligo)nucleotides, and fatty acids? Valuable insights into these questions have already been gained by combining theoretical and experimental approaches, for instance by computationally reproducing the famous Miller–Urey experiment,4−6 investigating ribose formation within the formose network,7,8 or studying noncatalytic oligomerization of ribonucleotides.9

Therefore, exploiting the predictive power of computational chemistry by exploring the intricate chemical reaction space of a given set of compounds is a captivating endeavor. However, computational modeling of chemical systems is also highly challenging due to the high spatial dimensionality and almost infinite possibilities for reaction channels and possible conformations. Improvements in both software and hardware, as well as in available algorithms have helped the field evolve to a fast-paced study area underpinned by several recent developments in the available methodology and scope of application.10

The common principle of all first-principles chemical reaction space exploration methods is very simple: exhaustively investigate the mechanistic paths available to a set of initial compounds by simulating the system’s dynamics on its potential energy surface (PES) and construct the corresponding reaction network connecting all found intermediates and products by adjacent reaction paths.11 While first attempts to explore the vastness of the chemical reaction space by computational means relied on reducing the multidimensional problem of chemical transformations to a two-dimensional matrix representation paired with heuristic concepts,12,13 recent methodology14−27 leverages the power of available data science approaches, such as machine learning and neural networks,28−30 as well as efficiently exploits modern computer hardware.31−37

Out of all available chemical reaction space exploration approaches, we will focus on a molecular dynamics (MD) based quantum-chemical first-principles method, the ab initio nanoreactor,6,19,38 first introduced by Wang et al., which uses a temperature-based approach. Herein, reactivity is enhanced by performing high-temperature simulations of a spherically confined molecular system. In addition to employing temperatures of up to a few thousand Kelvin, an external virtual piston ensures the periodical contraction of the nanoreactor sphere, which in turn increases pressure and enables bond cleavage and formation as investigated by Wolinski and Baker.39,40 As the exploration proceeds undirected, the ab initio nanoreactor is a forward open-end exploration method according to the categorization presented by Unsleber and Reiher.11 However, only small molecular species can be studied as larger molecules can easily fragment under the extreme conditions employed.8

Interestingly, a similar sampling challenge is posed in the biochemical context. Here, to accelerate the simulation of slow processes, such as protein folding, numerous so-called enhanced sampling techniques have been developed.41 These rely either on elevating the PES along a predefined reaction coordinate to decrease reaction barriers locally or modifying the canonical probability distribution of the whole system, eliminating the need for prior knowledge about the involved collective variables (CVs). The latter strategy is used in temperature-based methods like replica exchange42/parallel tempering43 (RE/PT) MD44 or integrated tempering sampling (ITS),45,46 as well as in bias potential-based methods, e.g., hyperdynamics,47,48 accelerated MD49 (aMD) and its successors, Gaussian accelerated MD50 (GaMD), and Sigmoid accelerated MD51 (SaMD). The hyperdynamics scheme by Voter et al.47,48 is based on identifying transition regions and elevating them by a boost potential dependent on the Hessian matrix and approximations thereof. Hamelberg et al.49 provides in aMD a different form for the bias potential to increase computational efficiency. Later, more general formulations for the bias potential were presented in GaMD50 and SaMD51 with the goal to improve sampling efficiency and accuracy.

Enhanced sampling methods using bias potentials along a predefined CV have already been successfully employed for automated chemical reaction space exploration as demonstrated by Grimme in his metadynamics-based approach, where the RMSD was employed as a CV.52 Herewith, valuable insights into the catalytic behavior of a P450 induced oxidation were achieved. Furthermore, Hirai and Jinnouchi used an adaptive form53 of the aMD method by Hamelberg et al.49 in combination with increased temperatures within the previously described nanoreactor framework to study the oxidation-induced decomposition of ethylene at 1273 K.54 They also extended their method to the treatment of solid state reactions,55 highlighting the potential of enhanced sampling methods to aid modeling.

Our aim is to extend our previous investigations on the ab initio nanoreactor approach, which involved the development of a new, milder form for the virtual piston, along with an automated simulation postprocessing workflow to highly speed up the generation of reaction networks from nanoreactor trajectories.8 For this purpose, we combine spherical confinement with hyperdynamics-inspired biasing strategies in an approach which we refer to as ab initio hyperreactor dynamics (HRD). This provides a robust method for conducting reaction space exploration under mild conditions.

Theoretical Background and Methods

Reaction Space Exploration With the Computational Nanoreactor

Based on the original nanoreactor developed by Martínez and co-workers,6 we have presented a systematic parameter study8 which also includes a comparison of different forms for the potential controlling the aforementioned virtual piston to assess its effect on the obtained reactivity and numerical stability of the simulation. In this context, we have introduced a new, milder form of the contracting spherical potential Vsphere

graphic file with name oc3c01403_m001.jpg 1

where N denotes the number of atoms. The magnitude of Vspheren acting on each atom n depends on the atomic radial coordinate rn, as well as on the imposed time-dependent radius for the nanoreactor sphere r0(t), to which the atoms are confined.

graphic file with name oc3c01403_m002.jpg 2
graphic file with name oc3c01403_m003.jpg 3

To ensure equal acceleration of all atoms at the same radial coordinate rn, the applied harmonic potential Vspheren is additionally mass-weighted by a unitless factor equal to the magnitude in atomic units (a.u.) of the corresponding atomic mass mn. The radius of confinement r0(t) oscillates according to eq 3 between rmin and rmax, following a smooth-step curve with a period of ttotal.8

This form for the temporal switch of the external potential and the corresponding spherical confinement has enabled increased stability of the performed simulations8 compared to the rectangular wave potential used in the original nanoreactor procedure.6 Therefore, the former was employed in the present study to induce pressure at low thermostat temperatures Ttarget along with the flattening of the potential energy surface by hyperdynamics-inspired methods49−51 in HRD, as presented in the following.

Enhanced Sampling by Hyperdynamics-Related Methods

The hyperdynamics method of Voter47,48 mentioned before was deemed unfeasible for enhancing the sampling of large systems due to the related increased computational cost. This led to the proposal of a trivial condition for applying a boost potential to lift the PES in minimum energy wells by Steiner et al.56 and its application to surface diffusion processes. Rahman and Tully57 showcased the effectivity of the latter method which they termed “puddle-skimming” for exploring the multidimensional configuration space by adding a boost potential ΔV(x) to the original potential V(x) (where xT = (x1, x2, x3, ..., x3N)) above a certain threshold value:

graphic file with name oc3c01403_m004.jpg 4

Herein, ΔV(x) = E – V(x) and E is the boost energy, below which the bias is applied to aid the escape from minimum energy wells. Therefore, the PES forms a “puddle” if V(x) < E where the molecular system experiences a random walk, as V*(x) = E. However, this simple definition for ΔV(x) leads to discontinuities in the first derivative of the modified potential at the points where V(x) = E and, therefore, requires a special integration technique to ensure temporal propagation on the PES during the MD.

Since then, several formulations for ΔV(x), which do not exhibit the above-mentioned discontinuities, have been proposed and applied to enhance the conformational sampling of biomolecules and enable the investigation of slow processes on computationally feasible time scales. Herein, we will focus on the aMD method49 and its successors, GaMD by Miao et al.,50 and the recent SaMD approach presented by Zhao et al.51

The formulations for ΔV(x) as introduced in eq 4 and its derivative with respect to V(x) for the three enhanced sampling methods mentioned above are summarized in Table 1. To ensure proper sampling during the simulation, the added potential should modify the PES in such a way that the underlying shape is preserved to a certain degree. The parameter E should be defined carefully, as if chosen below Vmin (the local minimum near the starting structure) the modified potential will always remain unbiased.

Table 1. Formulations of the Boost Potential ΔV and Its Derivative with Respect to V for aMD,49 GaMD,50 and SaMD51.

method ΔV(x)
graphic file with name oc3c01403_m006.jpg
aMD49
graphic file with name oc3c01403_m007.jpg
graphic file with name oc3c01403_m008.jpg
GaMD50
graphic file with name oc3c01403_m009.jpg
 –k(E – V(x))
SaMD51
graphic file with name oc3c01403_m010.jpg
graphic file with name oc3c01403_m011.jpg
Open in a new tab

In aMD, the parameter α defines the strength of the added bias, so that the modified potential transitions from a flat to the true PES for increasing values of α. Even though aMD provides a faster and more comprehensive method to explore the conformational and reaction space of large systems compared to conventional MD simulations, in many cases accurate free energy differences of the simulated processes are desired. To obtain the latter, the previously added bias must be removed. Unfortunately, due to the large boost potentials applied in aMD, which range between tens and hundreds of kilocalories per mole and also possess a wide distribution, this reweighting step has proven to be problematic and suffer from large energetic noise,58 which makes recovering the accurate free energy landscapes challenging.49

Therefore, GaMD aims to provide improved reweighting by using a harmonic boost potential ΔV(x) which follows near-Gaussian distribution.50 Finally, the recently introduced SaMD approach targets the improvement of sampling efficiency and effectivity of the applied boost potential while minimizing its standard deviation. As its name suggests, a sigmoid form was chosen for the first derivative of the boost potential (compare Table 1 third column, third row) with respect to V(x). For both GaMD and SaMD, the strength of the boost potential, and therefore of the applied bias, is controlled by a user-defined standard deviation σ0 for ΔV(x), which in turn determines k as described in the Supporting Information.

For better understanding the three hyperdynamics-inspired techniques, the different boost potentials ΔV(x) are plotted for a hypothetical potential on the left of Figure 1, along with the derivative with respect to V(x) of the corresponding bias force on the right. For all presented approaches, the boost energy E, along with the minimal and maximal potential energy, Vmin and Vmax, are estimated from the calculated potential during the equilibration simulation performed in the beginning.

Figure 1.

Figure 1

Open in a new tab

Boost potential applied to a hypothetical potential and derivative with respect to V(x) of corresponding bias forces for aMD (top), GaMD (middle), and SaMD (bottom).

Inducing Reactivity with Accelerated MD Methods

While both the computational nanoreactor method as introduced by Wang et al.6 and the adaptive aMD approach used by Hirai53 provide good results for small reactants and have proven to be of great use for simulating combustion, the extreme temperatures of thousands of Kelvin employed therein, which aid the reactivity by increasing the kinetic energy of the system, also unavoidably increase the risk for numerical instability and molecular fragmentation.

Our aim is to achieve similar reactive behavior as in the high-temperature approaches while keeping the target temperature of the thermostat low and transit to a more general reaction space exploration method. Hyperdynamics-related methods, such as aMD, GaMD, and SaMD provide useful tools for inducing reactive behavior without the need of choosing reaction coordinates a priori.

Figure 2 gives a quick and comprehensive overview of the workflow employed in accelerated HRD (aHRD), Gaussian accelerated HRD (GaHRD), and Sigmoid accelerated HRD (SaHRD) depending on the chosen boost potential ΔV(x). The definitions for V*(x) and Vspheren(rn, r0(t), kconf) are given in eq 2, eq 4, and Table 1. We expect this to enable investigation of a broader range of systems and extend the scope of the present temperature-based computational nanoreactor approach.

Figure 2.

Figure 2

Open in a new tab

Flowchart of the HRD procedure consisting of applying a periodic contraction on the molecules confined to a virtual nanoreactor sphere after having already biased the original potential using aMD, GaMD, or SaMD boost potentials. While for t ≤ tinit the original PES is preserved, the two bias potentials complement each other during the exploration phase after the equilibration of the chosen hyperdynamics procedure has been finished (t > teq). The boost potential ΔV(x) is applied in each MD step prior to the external harmonic potential Vsphere.

For the investigated systems, only selected products, intermediates, and reaction paths are included in the results. The focus of the present work has been on the exploration of chemical space and the comparison of the induced reactivity by the different hyperdynamics-inspired bias potentials. Therefore, our goal for the future is to enable (semi)automated refinement of the obtained reaction networks and paths of interests as already introduced by Martínez and co-workers59,60 for the conventional nanoreactor approach. For this purpose, we aim to leverage the ongoing efforts in our group to reduce computational scaling of accurate quantum chemical calculations, e.g., at the DFT level,31−33 to accelerate the refinement procedure.

Computational Details

All simulations included in this study were performed using the GFN2-xTB method61 for the computation of energies and gradients, whereby xtb(62) was interfaced to our in-house MD engine. Default settings were used for the SCF convergence criterion (ΔESCF ≤ 10–6Eh) and for the electronic temperature Tel = 300 K. The initialization in the nanoreactor sphere was done using preoptimized molecules at the PBEh-3c63/def2-mSVP level of theory using our initialization procedure introduced before.8

The implementation of aMD, GaMD, and SaMD boost potentials and forces was provided by the adaptive-sampling package,64,65 where the implementation was done as described in the original publications.49−51 The postprocessing of the simulations was fully automated by using our nanoreactor evaluation workflow.8,66

Simulation Details

For the method comparison of aHRD, GaHRD, and SaHRD, we have conducted triplicates for all tested parameters. For this purpose, a homogeneous test system of 50 HCN molecules enclosed in a theoretical nanoreactor sphere contracting between 15 and 7 Å was used and a Langevin thermostat at Ttarget = 298.15 K with a friction constant of γ = 7 ps–1 was employed for temperature control. For all simulations, 1 × 103 initial steps were used, while the number of equilibration steps was set to 1 × 104. The external spherical confinement was turned on after the equilibration was finished and σV, E, Vmin, and Vmax had been determined. The force constant k was computed for GaHRD and SaHRD as described in the Supporting Information. The interstellar low-temperature system consisting of acetaldehyde, ammonia, and radical counterparts was simulated at the reported67 experimental temperature of 10 K in vacuum. For the simulation of the nonenzymatic nucleoside synthesis, GaHRD and SaHRD simulations in vacuum and using implicit water solvation (ALPB model)168 at Ttarget = 323.15 K were compared to conventional nanoreactor simulations at Ttarget = 2000 K. Full details on all simulation parameters, as well as the initial geometries used for the HCN system, are given in the Supporting Information.

Results and Discussion

The newly developed HRD method is first tested extensively at Ttarget = 298.15 K on the literature-known8 HCN molecular setup consisting of 50 molecules (150 atoms), and the outcome is compared to the results obtained with the high-temperature ab initio nanoreactor procedure. Furthermore, to ensure good temperature and pressure control, we investigate the behavior of the HRD procedure at different equilibrium temperatures. The efficiency of HRD simulations is then showcased on two prebiotically relevant application setups, concerning the synthesis of indispensable molecular building blocks for the emergence of life.3

Exploring Reactivity Enhancement

To investigate the required strength for driving reaction space exploration by combining aMD, GaMD, or SaMD boost potentials with the periodic contracting external potential introduced in the context of the computational nanoreactor,6,8 a set of nine suitable acceleration parameters was determined for each HRD type by running unbiased MD simulations of the HCN test system described above and visualizing V*(x) against the unbiased potential energy surface. The aim was to test a wide range of acceleration parameters in terms of quantitative and qualitative reactivity outcome.

The chosen maximal trajectory length for the HRD exploration phase was 100 ps, including the equilibration period of 5 ps. Due to the varying trajectory lengths caused by different bias strength and resulting numerical problems, we report the number of new unique molecular species obtained every 2 ps, which corresponds to the period of the contracting spherical potential Vsphere. By applying this criterion, we aim to partially remove distortions caused by different simulation lengths.

The results obtained for aHRD, GaHRD, and SaHRD simulations of the test HCN system at 298.15 K and kconf = 1.00 kcal mol–1 Å–2 are shown in Figure 3 in orange. The achieved total simulation time is shown in dark gray for better interpretation of the results. An additional figure providing an enlargement for this confinement strength is given in Figure S2. We note that due to the computational effort needed to perform these simulations, we have restricted the number of parallel simulations to three per setup. Even though the results are not statistically converged and outliers were encountered, the performed simulations provide useful trends. The reactivity enhancement correlates as expected with the employed acceleration parameter α for aHRD and σ0 for GaHRD/SaHRD, as the number of obtained molecular species increases for decreasing α and increasing σ0, respectively. Furthermore, both reactivity and stability are clearly more sensitive to α than to σ0, which represents a user-defined upper threshold for the maximal standard deviation of the boost potential ΔV(x). Low values for α highly increase the reactivity; however, low stability of the simulation is observed, which leads to numerical problems. As expected from the behavior of Inline graphic, SaHRD exhibits the lowest sensitivity to the chosen σ0 by allowing a random walk in the flattened basins of the PES (compare Figure 1 c and Table 1).

Figure 3.

Figure 3

Open in a new tab

Results reported as number of new unique molecular species obtained every contraction–expansion period (2 ps) for HRD simulations employing different confinement force constants kconf and the nine selected acceleration parameters for each of the different HRD variants. The total simulation time was averaged for each acceleration parameter over four simulations employing different kconf. Low values for α induce great reactivity; however, they also lead to instability of the simulations and total simulation times of under 50 ps as highlighted in dark gray. The Gaussian accelerated approach provides the best balance between acceleration and stability by showing low sensitivity to the choice of kconf, while SaHRD delivers the best qualitative outcome overall (see Figure 4).

As we employ two reactivity-enhancement techniques for the HRD method, we have additionally performed calculations to show the interplay of the spherical confinement and the chosen strength of the hyper-MD bias. For this purpose, we compare the previously obtained results at kconf = 1.00 kcal mol–1 Å–2 to the outcome when gradually reducing the confinement strength to kconf = 0.25 kcal mol–1 Å–2. The results confirm earlier findings, that increasing the confinement strength in general aids the exploration process.8 However, it can also hinder it in the case of HRD when the two bias potentials become too strong together, leading to increased numerical problems and early termination of the simulations, or they cancel each other, therefore, hindering the exploration of new regions as shown in Figure 3. Considering these findings, we recommend using the highest confinement possible while applying a medium-strength acceleration parameter. This should provide a good balance between numerical stability of the simulation and induced reactivity.

From the three different tested HRD approaches, SaHRD performed best both quantitatively for the ideal kconf = 1.00 kcal mol–1 Å–2, as shown in Figure 3 in orange, and qualitatively by providing a wide range of compounds, including previously identified RNA precursors with the earlier piston-accelerated nanoreactor approach8 at 2000 K; e.g., methane (18), ethanedinitrile (19), formamidine (35), and cyanoacetylene (21) were found along with imidazole (36) when increasing the bias strength. A selection of identified compounds is summarized in Figure 4 based on the classification in primary and secondary RNA precursors presented by Benner et al.68 Selections for compounds obtained within aHRD and GaHRD simulations can be found in the Supporting Information, where an overview of numbered molecules obtained in all conducted HCN simulations within this work is also provided.

Figure 4.

Figure 4

Open in a new tab

Selection of obtained molecular species for SaHRD simulations at kconf = 1.00 kcal mol–1 Å–2 and rmin = 7 Å/rmax = 15 Å, ranging from simple carbohydrates to RNA precursors as postulated by Benner et al.,68 and important heterocyclic compounds, such as imidazole (36). Starting compounds are shown in green, while primary and secondary RNA precursors are depicted in orange and red, respectively.

Temperature and Pressure Regulation in Hyperreactor Dynamics Simulations

To check temperature and pressure regulation by the employed Langevin thermostat (Ttarget = 298.15 K; γ = 7 ps–1), we looked at the mean temperature and pressure throughout the simulation for every combination of kconf and α/σ0. Outliers were removed prior to the analysis by the IQR method.69 The results shown in Figure S3 reveal very good temperature control for most simulation setups except for two aHRD and two SaHRD cases, which are clearly outliers and caused by the very harsh bias employed. Furthermore, the pressure fluctuations have a clear tendency to increase for higher kconf, as expected and reported before,8 as well as for milder acceleration during the hyperdynamics. This finding can be attributed to increasing spherical confinement due to atom dissipation, and it systematically shows the interplay between the two bias potentials.

To check the performance of the HRD approach under varying equilibrium conditions, we have conducted investigations at four different temperatures, Ttarget = 10.00, 100.00, 273.15, and 323.15 K using the best performing setups from previous findings. Notably, the Langevin thermostat precisely regulates the temperature regardless of the employed acceleration parameter as shown on the left of Figure 5. This provides an excellent basis for conducting low temperature reaction space exploration studies.

Figure 5.

Figure 5

Open in a new tab

Investigations on HRD reaction space exploration at different potentially experimentally defined equilibrium temperatures. Excellent temperature regulation is achieved by the Langevin thermostat for very low temperatures down to 10 K. Reactivity is enhanced, as expected, by higher temperatures and acceleration strength. All simulations were conducted with kconf = 1.00 kcal mol–1 Å–2 and rmin = 7 Å/rmax = 15 Å.

In terms of obtained results, there is no clear correlation between the number of obtained unique molecular species every 2 ps and the employed equilibrium temperature paired with previously defined acceleration parameters for either aHRD, GaHRD, or SaHRD simulations. The reactivity tremendously increases regardless of the employed acceleration strength, when surpassing T = 100.00 K. We attribute this to the used test system, and it should not represent a general rule. Even though SaHRD shows low sensitivity to the choice of σ0 at T = 298.15 K in terms of simulation stability, the chosen temperature induces numerical problems when exceeding 300 K for the chosen test setup. Nevertheless, the obtained results were very satisfactory for all three methods and show that choosing the right temperature when aiming to observe certain chemical transformations using the HRD approach is crucial.

In addition, we have investigated the computational feasibility of aHRD, GaHRD, and SaHRD for the three best-performing acceleration parameters presented above compared to conventional nanoreactor simulations of the HCN system at 298.15 and 2000.00 K. The HRD simulations were conducted at an equilibrium temperature of 298.15 K. The results presented in Figure S1 reflect our previous observations regarding stability and reactivity of the simulations. The increased target temperature in the conventional nanoreactor approach leads to a computational overhead due to additional steps needed to heat and equilibrate the system prior to conducting the exploration phase, as well as plays a significant role in destabilizing the simulation and leading to numerical problems, therefore, also increasing the computation time. Furthermore, even though the type of the bias potential does not influence the performance, the suitability of the chosen acceleration parameter clearly does and the choice of the latter should therefore be made with care. On this note, we conclude that HRD delivers better performance for a wider variety of chemical systems than the conventional nanoreactor approach, as it allows for efficient reaction space exploration at low temperatures, therefore providing better numerical stability of the simulation and solving the problem of molecular fragmentation.

Application to Origins of Life Examples

To showcase the efficiency of HRD, two experimental setups were investigated at the experimental equilibrium temperatures.67,70 While the first system was chosen due to its extreme low temperature conditions, the second aimed to simulate reactions involving combined implicit-explicit solvation, more complex reactants prone to severe fragmentation, and a moderate experimental temperature of 50 °C.

Prebiotic Synthesis of Glycinal and Acetamide in Interstellar Ices

In a recent study,67 R. I. Kaiser and co-workers presented a novel possible interstellar access pathway for glycinal and acetamide initiated by simulated galactic cosmic rays in interstellar model ices at 10 K. These two molecular species represent important molecular building blocks for the formation of glycine units in prebiotic polymeric chains. The experimentally established procedure supports radical formation by irradiation of a mixture of acetaldehyde and ammonia leading to radical species by hydrogen elimination and subsequent recombination of the obtained radicals to glycinal, acetamide, N-methylformamide, and methyloxaziridine as shown in Figure 6 (upper right). These intermediate products can undergo barrierless recombination followed by rearrangement, ultimately yielding amide chains.

Figure 6.

Figure 6

Open in a new tab

Comparison of obtained molecular species and selected reactions in aHRD, GaHRD, and SaHRD simulations at T = 10 K (gray background) and experimentally obtained products on the upper right. The simulations were started from varying mixtures of acetaldehyde, ammonia, and radical counterparts shown in green. Obtained peptide precursors matching experimental observations67 are depicted in dark blue.

Given our preliminary investigations on the behavior of HRD at low equilibrium temperatures, we have conducted a study using different setups for the interstellar formation of glycinal, acetamide, and amides starting from acetaldehyde (42), ammonia (5), as well as including corresponding radical species (43, 44, 45). For this purpose, aHRD, GaHRD, and SaHRD simulations were performed at an equilibrium temperature of 10 K using different acceleration parameters. While for aHRD, α was determined through unbiased test simulations, for GaHRD and SaHRD we have decided to compare σ0 values ranging between 10 kBT, which represents the recommended upper threshold for correct reweighting, and 100 kBT. All further details on simulation parameters are given in the Supporting Information.

All tested setups support the experimental data with glycinal (46), acetamide (47), as well as radical counterparts (53, 56) being formed. New reaction paths potentially leading to glycinal and acetamide could be identified in all simulations regardless of the applied reactivity enhancement procedure. However, as already observed in the parameter screening using the homogeneous HCN setup, determining the right acceleration strength for aHRD is challenging and it highly relies on chemical intuition.

In addition to the two expected amide precursors mentioned before, 1-aminoethanol (74) and hemiacetals (85, 87) were formed in the radical-free simulations starting from a mixture of acetaldehyde and ammonia in a ratio of 13:14 (147 atoms). Herein, ammonia was often seen to act as a catalytic species.

To investigate radical reactions in this context, simulations of two systems involving different reactant-derived single radical species mixed with acetaldehyde and ammonia were conducted. Ratios of 6:1 and 1:1 of nonradical to radical species (shown in Figure 6 in green) were tested. Here, various reaction paths ranging from barrierless radical recombinations to nucleophilic additions and rearrangements involving heterocyclic species (51) could be identified. A wide variety of prebiotic compounds, such as hydrogen (4), nitrogen (3), methane (18), acetylene (14), hydrazine (63), water (61), carbon dioxide (64), methanol (65), and methylamine (40), was retrieved from the 300 ps-long simulations regardless of the employed acceleration parameters, merely the time of formation differing. Besides the obtained postulated precursors, namely glycinal, acetamide, and N-methylformamide (48) shown in dark blue in the lower left corner in Figure 6, numerous other reactive molecular species exhibiting various functional groups, such as alcohol, amine, (hemi)acetal, and carbonyl groups were obtained. These reactive compounds, among which glyoxal (81), 1-aminoethanol (74), and glycolaldehyde (76) were present, play a crucial role in driving reactivity during the exploration phase by providing protons and nucleophilic character.

Furthermore, carbamic acid (88) and urea (89), as well as aminoacetonitrile (26), which is a precursor of glycine, could be identified in the SaHRD simulations and provide a good molecular starting pool for the prebiotic nonenzymatic formation of amino acid chains. Overall, the performed HRD simulations were very efficient, GaHRD performing the best in terms of convergence and stability of the simulations which can be attributed to the biased potential preserving the topology of the PES. Regarding molecular variety, SaHRD performed best by additionally supporting the formation of longer carbon chains besides the expected products, therefore leading to species such as butanedial (82) and vinylacetate (86). However, in the case of SaHRD, increasing instability was encountered for simulations performed with stronger bias. We attribute this behavior to the flat topology of the modified PES at V(x) = E, as well as to the strong bias of up to σ0 = 100 kBT employed.

Three selected reaction paths leading to glycinal, acetamide, and N-methylformamide are shown in the lower right corner of Figure 6. While the former two compounds were obtained from acetaldehyde, ammonia, and a formylmethyl radical through nucleophilic attack at the radical species followed by hydrogen transfer, the synthesis of the latter involved a cyclization to 2-aziridinol (51) which enabled the rearrangement to N-methyleneformamide (52), and ultimately to N-methylformamide (48).

Nonenzymatic Synthesis of DNA Nucleosides

The way DNA formed on early Earth is still debated, as there is no conclusive evidence how this complex molecule, which plays an essential role for stable genetic information storage, could have formed from RNA in the absence of ribonucleotide reductases. Therefore, great effort has been put into finding synthesis pathways for deoxyribonucleotides under prebiotic conditions. In 2019, Trapp and co-workers70 presented a novel continuous path to deoxyribonucleosides starting from prebiotically available acetaldehyde (42) and d-glyceraldehyde (90) in an aqueous environment at 50 °C. The postulated reaction mechanism is summarized in the upper left of Figure 7 and involves an initial nucleophilic addition of acetaldehyde to N1/N9 of the pyrimidines and purines, respectively, followed by water elimination. The obtained enamine then initiates a nucleophilic attack at the carbonyl group and the compound finally cyclizes to yield the deoxyribonucleoside.

Figure 7.

Figure 7

Open in a new tab

Results of SaHRD simulations investigating a nonenzymatic DNA nucleoside synthesis (experimental results shown in the upper left corner) at T = 323.15 K and using σ0 = 1.023 × 10–2Eh (corresponds to 10 kBT). Reactants are given in green. Simulations were performed with and without the ALPB implicit solvent model for water, which is shown to support nucleophilic additions in this case as summarized in the upper right blue frame. For comparison, conventional nanoreactor simulations at T = 2000 K starting from the same initial geometries were conducted. However, these exhibit severe fragmentation of the initial reactants as shown in the lower left corner in black.

We have conducted investigations at Ttarget = 323.15 K starting from a 1:1:1:3 mixture of the nucleobase (A, G, T, C), acetaldehyde, d-glyceraldehyde, and water using the GaHRD and SaHRD approach with and without applying the ALPB implicit solvent model for water and compared the results to conventional computational nanoreactor simulations at 2000 K. Triplicates were performed for all simulations starting from three distinct initial configurations for each system. Full details on the employed parameters are given in the Supporting Information.

The obtained results clearly showcase the efficiency and advantages of HRD, as the nucleophilic addition of acetaldehyde at N9 leading to A1 was successfully reproduced for adenine, when using water as an implicit solvent. The use of implicit solvation and performing the simulations at Ttarget = 323.15 K supported nucleophilic additions for all nucleobases at various positions. Therefore, additions of acetaldehyde and glyceraldehyde were mainly seen at N3, N7, and N9 for adenine and guanine, and at N3 for cytosine. For thymine, only additions at the carbonyl oxygen were observed and the expected enamine could not be retrieved. The simulations performed in vacuum lacked the expected enamines completely and in this case, nucleophilic additions at N3 and N7, as well as directly at the carbonyl oxygen predominated and yielded precursors for further cyclizations as shown in the lower right corner of Figure 7. Given the obtained glyceraldehyde adducts, e.g., A4, A5, or G4, we believe that a sequential exploration phase, where d-glyceraldehyde is added after the enamine has already been formed to first increase the chance of collision of the nucleobase and acetaldehyde, is recommended, as the number of atoms in ab initio MD simulations is still limited by the efficiency of the employed electronic structure method, therefore also statistically influencing the outcome.

In contrast, the conventional computational nanoreactor approach6,8 at 2000 K resulted in severe fragmentation of the reactants to hydrogen peroxide (62), formic acid (94), formaldehyde (93), acetonitrile (20), glycolaldehyde (76), and substituted oxiranes (97) among many others. Furthermore, highly functional aliphatic chains (101, 102) were encountered. Only in one case, a nucleophilic addition was observed, at C8 of guanine, leading to G5.

Given the obtained results and the positive effect of implicit solvation and low thermostat temperatures on the studied system, the ab initio HRD approach delivers satisfactory outcomes at increased exploration rate and allows the use of experimentally defined reaction temperatures, as well as solvation environments for a better description of the molecular systems compared to the conventional ab initio nanoreactor method.

Conclusion

We have investigated alternative reactivity enhancement techniques in the context of the computational nanoreactor approach first introduced by Wang et al.6 in combination with the smooth-step regulated external potential developed in our group8 to enable reaction space exploration at lower thermostat temperatures. Therefore, we have combined hyperdynamics approaches, namely aMD,49 GaMD,50 and SaMD,51 to elevate the PES with the smooth-step driven virtual piston to keep the molecules confined to the nanoreactor sphere within a new reaction space exploration approach termed ab initio hyperreactor dynamics (HRD).

Investigations on the optimal choice of the acceleration parameter (α or σ0) for a toy molecular system composed of 50 HCN molecules show positive correlation between the strength of the chosen boost potential ΔV(x) and the observed reactivity. Furthermore, the interplay between the strength of the periodic external potential given by kconf and α/σ0 was also carefully examined to better understand the effect of the combined biases on the outcome. We recommend choosing kconf between 0.75 and 1.00 kcal mol–1 Å–2 to achieve good confinement in the case of GaHRD/SaHRD at the recommended value for σ050 of 10 kBT and T = 298.15 K. The choice of an ideal α in aHRD has proven to be difficult because it depends on the magnitude of the system’s initial potential energy.

As one of our main goals was to enable efficient and automated reaction space exploration at low equilibrium temperatures in order to avoid numerical problems and nonphysical structures, good temperature regulation was also ensured at the chosen friction constant γ = 7 ps–1 regardless of the chosen acceleration parameter and kconf. In addition, the influence of the thermostat temperature on the obtained reactivity for the three different acceleration approaches was investigated and revealed as expected a general increase in reactivity for higher equilibrium temperatures along with increased instability when surpassing T = 300 K. However, for all parameter investigations, the obtained results exhibit high variance depending on the initial configuration of the system, which makes averaging over multiple simulations employing the same parameters and different initial arrangements a necessity. Furthermore, out of the three tested hyperdynamics approaches, GaHRD provides increased stability independent of the chosen kconf, while a higher molecular variety is obtained with SaHRD.

To examine the performance of the HRD approach, we have chosen two experimentally investigated molecular systems stemming from origin of life hypotheses: the formation of glycinal and acetamide as precursors for polypeptides in an extraterrestrial synthesis67 at 10 K and a nonenzymatic synthesis of DNA nucleosides70 at 323.15 K. In the former case, we report the formation of the postulated precursors, glycinal, acetamide, and N-methylformamide, through various reaction pathways along with a multitude of prebiotically relevant molecules, which could aid the formation of more complex species under the experimentally described conditions. By including implicit water solvation within the ALPB model to the system composed of nucleobases, acetaldehyde, d-glyceraldehyde, and water necessary to model the nonenzymatic synthesis of DNA nucleosides, the postulated addition of acetaldehyde at N9 of adenine was achieved consistently. Additionally, the use of the HRD method supported nucleophilic additions at various positions of the nucleobases and hindered severe fragmentation of the reactants as observed in conventional computational nanoreactor simulations at 2000 K for the same molecular setups.

In conclusion, we report improved efficiency of the exploration phase within the computational nanoreactor framework by replacing the high thermostat temperatures by hyperdynamics derived bias potentials. The latter support the exploration of novel reaction pathways on an elevated PES with (partially) preserved topology under mild conditions. The presented approach represents a promising first-principles reaction space exploration method for modeling a wide variety of systems, regardless of the complexity of the reactants by employing a faster exploration protocol for the ideal acceleration strength than the conventional nanoreactor approach as demonstrated for the HCN system. Compared to the previously introduced metadynamics-based exploration protocol by Grimme,52 HRD enables an undirected exploration by using a CV-free enhanced sampling approach, therefore allowing for resampling of already visited regions on the PES. In the future, the method should benefit from the development of an adaptive version of the bias potential as already introduced and demonstrated for aMD,54 as well as from further automatization of the parameter choice.

Acknowledgments

The authors thank J. Kussmann (LMU Munich) for providing a development version of the FermiONs++ program package. The authors acknowledge financial support by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - Project-ID 364653263 - TRR 235 “Emergence of Life” and by the “Deutsche Forschungsgemeinschaft” (DFG, German Research Foundation) - SFB 1309-325871075 “Chemical Biology of Epigenetic Modifications”. C.O. acknowledges further support as Max-Planck-Fellow at the MPI-FKF Stuttgart.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscentsci.3c01403.

  • Equations for determining k in GaMD and SaMD, details on the nanoreactor-processing package used for postprocessing of the simulations, all simulation parameters, additional figures on computational timings, obtained molecular species for all systems, and temperature and pressure regulation during the HCN test simulations, as well as initial geometries for the HCN system (PDF)

The authors declare no competing financial interest.

Supplementary Material

oc3c01403_si_001.pdf (901.1KB, pdf)

References

  1. Keith J. A.; Vassilev-Galindo V.; Cheng B.; Chmiela S.; Gastegger M.; Müller K. R.; Tkatchenko A. Combining Machine Learning and Computational Chemistry for Predictive Insights into Chemical Systems. Chem. Rev. 2021, 121, 9816–9872. 10.1021/acs.chemrev.1c00107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Stiriba S. E. Computational Chemistry and Catalysis: Prediction and Design. Catalysts 2023, 13, 839. 10.3390/catal13050839. [DOI] [Google Scholar]
  3. Kitadai N.; Maruyama S. Origins of building blocks of life: A review. Geosci. Front. 2018, 9, 1117–1153. 10.1016/j.gsf.2017.07.007. [DOI] [Google Scholar]
  4. Miller S. L. A production of amino acids under possible primitive earth conditions. Science 1953, 117, 528–529. 10.1126/science.117.3046.528. [DOI] [PubMed] [Google Scholar]
  5. Miller S. L.; Urey H. C. Organic compound synthesis on the primitive earth. Science 1959, 130, 245–251. 10.1126/science.130.3370.245. [DOI] [PubMed] [Google Scholar]
  6. Wang L. P.; Titov A.; McGibbon R.; Liu F.; Pande V. S.; Martínez T. J. Discovering chemistry with an ab initio nanoreactor. Nat. Chem. 2014, 6, 1044–1048. 10.1038/nchem.2099. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Haas M.; Lamour S.; Christ S. B.; Trapp O. Mineral-mediated carbohydrate synthesis by mechanical forces in a primordial geochemical setting. Commun. Chem. 2020, 3, 1–6. 10.1038/s42004-020-00387-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Stan A.; von der Esch B.; Ochsenfeld C. Fully Automated Generation of Prebiotically Relevant Reaction Networks from Optimized Nanoreactor Simulations. J. Chem. Theory Comput. 2022, 18, 6700–6712. 10.1021/acs.jctc.2c00754. [DOI] [PubMed] [Google Scholar]
  9. Dass A. V.; Wunnava S.; Langlais J.; von der Esch B.; Krusche M.; Ufer L.; Chrisam N.; Dubini R. C. A.; Gartner F.; Angerpointner S.; Dirscherl C. F.; Rovó P.; Mast C. B.; Šponer J. E.; Ochsenfeld C.; Frey E.; Braun D. RNA Oligomerisation without Added Catalyst from 2′,3′-Cyclic Nucleotides by Drying at Air-Water Interfaces. ChemSystemsChem. 2023, 5, e20220002. 10.1002/syst.202200026. [DOI] [Google Scholar]
  10. Dewyer A. L.; Argüelles A. J.; Zimmerman P. M. Methods for exploring reaction space in molecular systems. Wiley Interdiscip. Rev. Comput. Mol. Sci. 2018, 8, e1354. 10.1002/wcms.1354. [DOI] [Google Scholar]
  11. Unsleber J. P.; Reiher M. The Exploration of Chemical Reaction Networks. Annu. Rev. Phys. Chem. 2020, 71, 121–163. 10.1146/annurev-physchem-071119-040123. [DOI] [PubMed] [Google Scholar]
  12. Broadbelt L. J.; Stark S. M.; Klein M. T. Computer Generated Pyrolysis Modeling: On-the-Fly Generation of Species, Reactions, and Rates. Ind. Eng. Chem. Res. 1994, 33, 790–799. 10.1021/ie00028a003. [DOI] [Google Scholar]
  13. Broadbelt L. J.; Stark S. M.; Klein M. T. Computer generated reaction modelling: Decomposition and encoding algorithms for determining species uniqueness. Comput. Chem. Eng. 1996, 20, 113–129. 10.1016/0098-1354(94)00009-D. [DOI] [Google Scholar]
  14. Maeda S.; Morokuma K. Finding reaction pathways of type A + B → X: Toward systematic prediction of reaction mechanisms. J. Chem. Theory Comput. 2011, 7, 2335–2345. 10.1021/ct200290m. [DOI] [PubMed] [Google Scholar]
  15. Maeda S.; Harabuchi Y. Exploring paths of chemical transformations in molecular and periodic systems: An approach utilizing force. Wiley Interdiscip. Rev. Comput. Mol. Sci. 2021, 11, e1538. 10.1002/wcms.1538. [DOI] [Google Scholar]
  16. Pieri E.; Lahana D.; Chang A. M.; Aldaz C. R.; Thompson K. C.; Martínez T. J. The non-adiabatic nanoreactor: Towards the automated discovery of photochemistry. Chem. Sci. 2021, 12, 7294–7307. 10.1039/D1SC00775K. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Martínez-Núñez E.; Barnes G. L.; Glowacki D. R.; Kopec S.; Peláez D.; Rodríguez A.; Rodríguez-Fernández R.; Shannon R. J.; Stewart J. J.; Tahoces P. G.; et al. AutoMeKin2021: An open-source program for automated reaction discovery. J. Comput. Chem. 2021, 42, 2036–2048. 10.1002/jcc.26734. [DOI] [PubMed] [Google Scholar]
  18. Chen X.; Liu M.; Gao J. CARNOT: A Fragment-Based Direct Molecular Dynamics and Virtual-Reality Simulation Package for Reactive Systems. J. Chem. Theory Comput. 2022, 18, 1297–1313. 10.1021/acs.jctc.1c01032. [DOI] [PubMed] [Google Scholar]
  19. Xu R.; Meisner J.; Chang A. M.; Thompson K. C.; Martínez T. J. First principles reaction discovery: from the Schrodinger equation to experimental prediction for methane pyrolysis. Chem. Sci. 2023, 14, 7447. 10.1039/D3SC01202F. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Nishimura Y.; Nakai H. Species-selective nanoreactor molecular dynamics simulations based on linear-scaling tight-binding quantum chemical calculations. J. Chem. Phys. 2023, 158, 54106. 10.1063/5.0132573. [DOI] [PubMed] [Google Scholar]
  21. Unsleber J. P.; Grimmel S. A.; Reiher M. Chemoton 2.0: Autonomous Exploration of Chemical Reaction Networks. J. Chem. Theory Comput. 2022, 18, 5393–5409. 10.1021/acs.jctc.2c00193. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Unsleber J. P. Accelerating Reaction Network Explorations with Automated Reaction Template Extraction and Application. J. Chem. Inf. Model. 2023, 63, 3392–3403. 10.1021/acs.jcim.3c00102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Unsleber J. P.; Liu H.; Talirz L.; Weymuth T.; Mörchen M.; Grofe A.; Wecker D.; Stein C. J.; Panyala A.; Peng B.; Kowalski K.; Troyer M.; Reiher M. High-throughput ab initio reaction mechanism exploration in the cloud with automated multi-reference validation. J. Chem. Phys. 2023, 158, 084803. 10.1063/5.0136526. [DOI] [PubMed] [Google Scholar]
  24. Ismail I.; Chantreau Majerus R.; Habershon S. Graph-Driven Reaction Discovery: Progress, Challenges, and Future Opportunities. J. Phys. Chem. A 2022, 126, 7051–7069. 10.1021/acs.jpca.2c06408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Hashemi A.; Bougueroua S.; Gaigeot M. P.; Pidko E. A. ReNeGate: A Reaction Network Graph-Theoretical Tool for Automated Mechanistic Studies in Computational Homogeneous Catalysis. J. Chem. Theory Comput. 2022, 18, 7470–7482. 10.1021/acs.jctc.2c00404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Türtscher P. L.; Reiher M. PATHFINDER–Navigating and Analyzing Chemical Reaction Networks with an Efficient Graph-Based Approach. J. Chem. Inf. Model. 2023, 63, 147–160. 10.1021/acs.jcim.2c01136. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Ramos-Sánchez P.; Harvey J. N.; Gámez J. A. An automated method for graph-based chemical space exploration and transition state finding. J. Comput. Chem. 2023, 44, 27–42. 10.1002/jcc.27011. [DOI] [PubMed] [Google Scholar]
  28. Stocker S.; Csányi G.; Reuter K.; Margraf J. T. Machine learning in chemical reaction space. Nat. Commun. 2020, 11, 5505. 10.1038/s41467-020-19267-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Schwaller P.; Vaucher A. C.; Laplaza R.; Bunne C.; Krause A.; Corminboeuf C.; Laino T. Machine intelligence for chemical reaction space. Wiley Interdiscip. Rev. Comput. Mol. Sci. 2022, 12, e1604. 10.1002/wcms.1604. [DOI] [Google Scholar]
  30. Khalak Y.; Tresadern G.; Hahn D. F.; de Groot B. L.; Gapsys V. Chemical Space Exploration with Active Learning and Alchemical Free Energies. J. Chem. Theory Comput. 2022, 18, 6259–6270. 10.1021/acs.jctc.2c00752. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Kussmann J.; Ochsenfeld C. Preselective Screening for Linear-Scaling Exact Exchange-Gradient Calculations for Graphics Processing Units and General Strong-Scaling Massively Parallel Calculations. J. Chem. Theory Comput. 2015, 11, 918–922. 10.1021/ct501189u. [DOI] [PubMed] [Google Scholar]
  32. Kussmann J.; Ochsenfeld C. Hybrid CPU/GPU Integral Engine for Strong-Scaling Ab Initio Methods. J. Chem. Theory Comput. 2017, 13, 3153–3159. 10.1021/acs.jctc.6b01166. [DOI] [PubMed] [Google Scholar]
  33. Kussmann J.; Laqua H.; Ochsenfeld C. Highly Efficient Resolution-of-Identity Density Functional Theory Calculations on Central and Graphics Processing Units. J. Chem. Theory Comput. 2021, 17, 1512–1521. 10.1021/acs.jctc.0c01252. [DOI] [PubMed] [Google Scholar]
  34. Laqua H.; Thompson T. H.; Kussmann J.; Ochsenfeld C. Highly Efficient, Linear-Scaling Seminumerical Exact-Exchange Method for Graphic Processing Units. J. Chem. Theory Comput. 2020, 16, 1456–1468. 10.1021/acs.jctc.9b00860. [DOI] [PubMed] [Google Scholar]
  35. Laqua H.; Kussmann J.; Ochsenfeld C. Accelerating semi-numerical Fock-exchange calculations using mixed single-and double-precision arithmetic. J. Chem. Phys. 2021, 154, 214116. 10.1063/5.0045084. [DOI] [PubMed] [Google Scholar]
  36. Laqua H.; Dietschreit J. C.; Kussmann J.; Ochsenfeld C. Accelerating Hybrid Density Functional Theory Molecular Dynamics Simulations by Seminumerical Integration, Resolution-of-the-Identity Approximation, and Graphics Processing Units. J. Chem. Theory Comput. 2022, 18, 6010–6020. 10.1021/acs.jctc.2c00509. [DOI] [PubMed] [Google Scholar]
  37. Seritan S.; Bannwarth C.; Fales B. S.; Hohenstein E. G.; Isborn C. M.; Kokkila-Schumacher S. I. L.; Li X.; Liu F.; Luehr N.; Snyder J. W.; Song C.; Titov A. V.; Ufimtsev I. S.; Wang L.-P.; Martínez T. J. TeraChem: A graphical processing unit-accelerated electronic structure package for large-scale ab initio molecular dynamics. Wiley Interdiscip. Rev. Comput. Mol. Sci. 2021, 11, e1494. 10.1002/wcms.1494. [DOI] [Google Scholar]
  38. Hutchings M.; Liu J.; Qiu Y.; Song C.; Wang L. P. Bond-Order Time Series Analysis for Detecting Reaction Events in Ab Initio Molecular Dynamics Simulations. J. Chem. Theory Comput. 2020, 16, 1606–1617. 10.1021/acs.jctc.9b01039. [DOI] [PubMed] [Google Scholar]
  39. Wolinski K.; Baker J. Theoretical predictions of enforced structural changes in molecules. Mol. Phys. 2009, 107, 2403–2417. 10.1080/00268970903321348. [DOI] [Google Scholar]
  40. Wolinski K.; Baker J. Geometry optimization in the presence of external forces: a theoretical model for enforced structural changes in molecules. Mol. Phys. 2010, 108, 1845–1856. 10.1080/00268976.2010.492795. [DOI] [Google Scholar]
  41. Yang Y. I.; Shao Q.; Zhang J.; Yang L.; Gao Y. Q. Enhanced sampling in molecular dynamics. J. Chem. Phys. 2019, 151, 70902. 10.1063/1.5109531. [DOI] [PubMed] [Google Scholar]
  42. Sugita Y.; Okamoto Y. Replica-exchange molecular dynamics method for protein folding. Chem. Phys. Lett. 1999, 314, 141–151. 10.1016/S0009-2614(99)01123-9. [DOI] [Google Scholar]
  43. Hansmann U. H. E. Parallel tempering algorithm for conformational studies of biological molecules. Chem. Phys. Lett. 1997, 281, 140–150. 10.1016/S0009-2614(97)01198-6. [DOI] [Google Scholar]
  44. Okamoto Y. Generalized-ensemble algorithms: enhanced sampling techniques for Monte Carlo and molecular dynamics simulations. J. Mol. Graph. Model. 2004, 22, 425–439. 10.1016/j.jmgm.2003.12.009. [DOI] [PubMed] [Google Scholar]
  45. Gao Y. Q. An integrate-over-temperature approach for enhanced sampling. J. Chem. Phys. 2008, 128, 064105. 10.1063/1.2825614. [DOI] [PubMed] [Google Scholar]
  46. Yang L.; Liu C. E.; Shao Q.; Zhang J.; Gao Y. Q. From thermodynamics to kinetics: Enhanced sampling of rare events. Acc. Chem. Res. 2015, 48, 947–955. 10.1021/ar500267n. [DOI] [PubMed] [Google Scholar]
  47. Voter A. F. A method for accelerating the molecular dynamics simulation of infrequent events. J. Chem. Phys. 1997, 106, 4655–4677. 10.1063/1.473503. [DOI] [Google Scholar]
  48. Voter A. F. Hyperdynamics: Accelerated Molecular Dynamics of Infrequent Events. Phys. Rev. Lett. 1997, 78, 3908–3911. 10.1103/PhysRevLett.78.3908. [DOI] [Google Scholar]
  49. Hamelberg D.; Mongan J.; Mccammon J. A. Accelerated molecular dynamics: A promising and efficient simulation method for biomolecules. J. Chem. Phys. 2004, 120, 11919. 10.1063/1.1755656. [DOI] [PubMed] [Google Scholar]
  50. Miao Y.; Feher V. A.; McCammon J. A. Gaussian Accelerated Molecular Dynamics: Unconstrained Enhanced Sampling and Free Energy Calculation. J. Chem. Theory Comput. 2015, 11, 3584–3595. 10.1021/acs.jctc.5b00436. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Zhao Y.; Zhang J.; Zhang H.; Gu S.; Deng Y.; Tu Y.; Hou T.; Kang Y. Sigmoid Accelerated Molecular Dynamics: An Efficient Enhanced Sampling Method for Biosystems. J. Phys. Chem. Lett. 2023, 14, 1103–1112. 10.1021/acs.jpclett.2c03688. [DOI] [PubMed] [Google Scholar]
  52. Grimme S. Exploration of Chemical Compound, Conformer, and Reaction Space with Meta-Dynamics Simulations Based on Tight-Binding Quantum Chemical Calculations. J. Chem. Theory Comput. 2019, 15, 2847–2862. 10.1021/acs.jctc.9b00143. [DOI] [PubMed] [Google Scholar]
  53. Hirai H. Practical hyperdynamics method for systems with large changes in potential energy. J. Chem. Phys. 2014, 141, 234109. 10.1063/1.4903787. [DOI] [PubMed] [Google Scholar]
  54. Hirai H.; Jinnouchi R. Discovering chemical reaction pathways using accelerated molecular dynamics simulations and network analysis tools – Application to oxidation induced decomposition of ethylene carbonate. Chem. Phys. Lett. 2021, 770, 138439. 10.1016/j.cplett.2021.138439. [DOI] [Google Scholar]
  55. Hirai H.; Jinnouchi R. Discovering surface reaction pathways using accelerated molecular dynamics and network analysis tools. RSC Adv. 2022, 12, 23274. 10.1039/D2RA04343B. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Steiner M. M.; Genilloud P.-A.; Wilkins J. W. Simple bias potential for boosting molecular dynamics with the hyperdynamics scheme. Phys. Rev. B 1998, 57, 10236–10239. 10.1103/PhysRevB.57.10236. [DOI] [Google Scholar]
  57. Rahman J. A.; Tully J. C. Puddle-skimming: An efficient sampling of multidimensional configuration space. Appl. Phys. Lett. 2002, 116, 8750–8760. 10.1063/1.1469605. [DOI] [Google Scholar]
  58. Shen T.; Hamelberg D. A statistical analysis of the precision of reweighting-based simulations. J. Chem. Phys. 2008, 129, 40. 10.1063/1.2944250. [DOI] [PubMed] [Google Scholar]
  59. Wang L. P.; McGibbon R. T.; Pande V. S.; Martínez T. J. Automated Discovery and Refinement of Reactive Molecular Dynamics Pathways. J. Chem. Theory Comput. 2016, 12, 638–649. 10.1021/acs.jctc.5b00830. [DOI] [PubMed] [Google Scholar]
  60. Chang A. M.; Meisner J.; Xu R.; Martínez T. J. Efficient Acceleration of Reaction Discovery in the Ab Initio Nanoreactor: Phenyl Radical Oxidation Chemistry. J. Phys. Chem. A 2023, 127, 9580–9589. 10.1021/acs.jpca.3c05484. [DOI] [PubMed] [Google Scholar]
  61. Bannwarth C.; Ehlert S.; Grimme S. GFN2-xTB - An Accurate and Broadly Parametrized Self-Consistent Tight-Binding Quantum Chemical Method with Multipole Electrostatics and Density-Dependent Dispersion Contributions. J. Chem. Theory Comput. 2019, 15, 1652–1671. 10.1021/acs.jctc.8b01176. [DOI] [PubMed] [Google Scholar]
  62. Bannwarth C.; Caldeweyher E.; Ehlert S.; Hansen A.; Pracht P.; Seibert J.; Spicher S.; Grimme S. Extended tight-binding quantum chemistry methods. Wiley Interdiscip. Rev. Comput. Mol. Sci. 2021, 11, e1493. 10.1002/wcms.1493. [DOI] [Google Scholar]
  63. Grimme S.; Brandenburg J. G.; Bannwarth C.; Hansen A. Consistent structures and interactions by density functional theory with small atomic orbital basis sets. J. Chem. Phys. 2015, 143, 54107. 10.1063/1.4927476. [DOI] [PubMed] [Google Scholar]
  64. Hulm A.; Dietschreit J. C. B.; Ochsenfeld C. Statistically optimal analysis of the extended-system adaptive biasing force (eABF) method. J. Chem. Phys. 2022, 157, 024110. 10.1063/5.0095554. [DOI] [PubMed] [Google Scholar]
  65. Hulm A.; Lemke Y.; Dietschreit J.; Glinkina L.; Stan-Bernhardt A.. Adaptive Sampling. https://github.com/ochsenfeld-lab/adaptive_sampling (accessed 2023-10-01).
  66. Stan-Bernhardt A.; Glinkina L.. Processing of Computational Nanoreactor Simulations. https://github.com/ochsenfeld-lab/nanoreactor_processing (accessed 2023-10-01).
  67. Marks J. H.; Wang J.; Kleimeier N. F.; Turner A. M.; Eckhardt A. K.; Kaiser R. I. Prebiotic Synthesis and Isomerization in Interstellar Analog Ice: Glycinal, Acetamide, and Their Enol Tautomers. Angew. Chem., Int. Ed. 2023, 62, e202218645. 10.1002/anie.202218645. [DOI] [PubMed] [Google Scholar]
  68. Ehlert S.; Stahn M.; Spicher S.; Grimme S. Robust and Efficient Implicit Solvation Model for Fast Semiempirical Methods. J. Chem. Theory Comput. 2021, 17, 4250–4261. 10.1021/acs.jctc.1c00471. [DOI] [PubMed] [Google Scholar]
  69. Benner S. A.; Bell E. A.; Biondi E.; Brasser R.; Carell T.; Kim H.; Mojzsis S. J.; Omran A.; Pasek M. A.; Trail D. When Did Life Likely Emerge on Earth in an RNA-First Process?. ChemSystemsChem. 2020, 2, e1900035. 10.1002/syst.202000009. [DOI] [Google Scholar]
  70. Haslwanter T.An Introduction to Statistics with Python, 1st ed.; Springer: Cham, 2016. [Google Scholar]
  71. Teichert J. S.; Kruse F. M.; Trapp O. Direct Prebiotic Pathway to DNA Nucleosides. Angew. Chem., Int. Ed. 2019, 58, 9944–9947. 10.1002/anie.201903400. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

oc3c01403_si_001.pdf (901.1KB, pdf)

Articles from ACS Central Science are provided here courtesy of American Chemical Society

RESOURCES