Combining Machine Learning Approaches and Accurate Ab Initio Enhanced Sampling Methods for Prebiotic Chemical Reactions in Solution

Abstract

The study of the thermodynamics, kinetics, and microscopic mechanisms of chemical reactions in solution requires the use of advanced free-energy methods for predictions to be quantitative. This task is however a formidable one for atomistic simulation methods, as the cost of quantum-based ab initio approaches, to obtain statistically meaningful samplings of the relevant chemical spaces and networks, becomes exceedingly heavy. In this work, we critically assess the optimal structure and minimal size of an ab initio training set able to lead to accurate free-energy profiles sampled with neural network potentials. The results allow one to propose an ab initio protocol where the ad hoc inclusion of a machine-learning (ML)-based task can significantly increase the computational efficiency, while keeping the ab initio accuracy and, at the same time, avoiding some of the notorious extrapolation risks in typical atomistic ML approaches. We focus on two representative, and computationally challenging, reaction steps of the classic Strecker-cyanohydrin mechanism for glycine synthesis in water solution, where the main precursors are formaldehyde and hydrogen cyanide. We demonstrate that indistinguishable ab initio quality results are obtained, thanks to the ML subprotocol, at about 1 order of magnitude less of computational load.

Non-linear temperature dependence of nitrogen adsorption and decomposition on Fe(111) surface

“…A satisfactory compromise between accuracy and efficiency can be achieved if one follows the pioneering work of Behler and Parrinello [27] and trains a machine learning potential (MLP) to reproduce a suitably chosen set of quantum mechanical calculations. By combining these machine learning techniques with advanced sampling methods, this approach has been shown to reproduce well the potential energy surface of different reactive systems [28][29][30][31][32][33][34][35][36][37].…”

Modeling Exchange Reactions in Covalent Adaptable Networks with Machine Learning Force Fields

“…The potential energy E of each configuration is the sum of individual atomic energies E = prefix∑ i E i , where E i is the energy contribution of a single atom determined by its local atomic environment. In this work, a cutoff of 6.0 Å was used to determine the local atomic environment, which is a common setting for the DPMD scheme, − and a smooth function was applied from 5.0 Å to ensure the components decay to zero smoothly at the boundary of cutoff.…”

“…However, the utilization of high-dimensional CVs s and z may increase the risk of hysteresis and slow convergence in calculating the free energy landscape, since multiple coordination numbers are used for the definition of CVs and a large CV space volume needs to be explored . To overcome this obstacle and obtain a converged free energy landscape, we performed committor analysis and umbrella sampling after the preliminary exploration of the free energy landscape using metadynamics. ,, For the reaction from reactant to intermediate of the amine-imine exchange, we extracted atomic configurations close to reactive events in the metadynamics trajectories and generated 100 independent unbiased MD trajectories with different initial random velocities starting from each configuration, verifying whether the trajectory falls into the reactant or intermediate basin. A transition state configuration was identified if the committor trajectories fell into both basins with a probability of 50 ± 10%.…”

“…Metadynamics was first employed to obtain a preliminary estimation of the free energy landscape for the aqueous model. 34 Figure 6a presents the evolution of CV s as a function of simulation time. It is seen that the trajectory explores three forward and two backward reactions within 5 ns.…”

“…63 The purpose of metadynamics simulations in this study is to obtain an estimation of the free energy landscape and get possible reaction pathways rather than construct an exact free energy landscape. To construct an exact free energy profile, we carried out committor analyses and validated the transition state and reaction pathway 34,58 as indicated in Figure S4a.…”