An automated method for graph‐based chemical space exploration and transition state finding

Abstract

Algorithms that automatically explore the chemical space have been limited to chemical systems with a low number of atoms due to expensive involved quantum calculations and the large amount of possible reaction pathways. The method described here presents a novel solution to the problem of chemical exploration by generating reaction networks with heuristics based on chemical theory. First, a second version of the reaction network is determined through molecular graph transformations acting upon functional groups of the reacting. Only transformations that break two chemical bonds and form two new ones are considered, leading to a significant performance enhancement compared to previously presented algorithm. Second, energy barriers for this reaction network are estimated through quantum chemical calculations by a growing string method, which can also identify non-octet species missed during the previous step and further define the reaction network. The proposed algorithm has been successfully applied to five different chemical reactions, in all cases identifying the most important reaction pathways.

Uncertainty quantification for predictions of atomistic neural networks

“…where y i is the i-th prediction and ŷt is the average value of the prediction in a bin B j . Using eqn (10) and (11), the Expected Normalized Calibration Error (ENCE):…”

“…In chemistry, such ML models are used in various branches ranging from the study of reactive processes, 1,2 sampling equilibrium states, 3 the generation of accurate force elds, [4][5][6][7][8] to the generation and exploration of chemical space. [9][10][11] Nowadays, an extensive range of robust and complex models can be found. [12][13][14][15][16] The quality of these models is only limited by the quality and quantity of the data used for training.…”

Exploring Chemical Space Using Ab Initio Hyperreactor Dynamics

“…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 . 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, , recent methodology − leverages the power of available data science approaches, such as machine learning and neural networks, − as well as efficiently exploits modern computer hardware. − …”

Diffusion-based Generative AI for Exploring Transition States from 2D Molecular Graphs

“…Both the conventional and ML approaches need an appropriate input preparation for 3D molecular geometries. However, it is well known that the results of the conventional approaches are sensitive to the input structures [33,34,55,56]. The ML approaches also take the 3D conformations of reactants and products as input.…”