Advancing Physical Chemistry with Machine Learning

A review on machine learning algorithms for the ionic liquid chemical space

“…Fast forward to the 21 st Century, and the evolution of the calculation power of modern computing systems has given machine learning methods (ML) the capacity to perform complicated calculations with extreme time and resources efficiency which are being used by major technological companies. 15 There are many detailed manuscripts on the history and evolution of ML, some indicative works are cited here. 16,17 ML methods are currently being implemented in research in a wide range of scientic elds, including chemical discovery and molecular design.…”

Infrared Spectroscopy of Chemically Diverse Carbon Clusters: A Data-Driven Approach

“…However, extracting the spectral features from a statistical sample seems also naturally suited to be tackled by machine learning (ML) techniques, a broad range of computational approaches that have become spectacularly popular in chemical physics and physical chemistry in the recent years. 15,16 Within the general context of relating properties to structure, several groups have recently shown the benefits of employing ML for vibrational spectroscopy [17][18][19] through a variety of approaches aiming to represent potential energy and electric dipole moment surfaces within perturbative frameworks, 20 to condense molecular information into topological descriptors, 21 or to numerically solve the quantum nuclear dynamics problem by better partitioning the various degrees of freedom. 22 In the present contribution, we explore several ML ideas to reconstruct the infrared spectrum of carbon clusters in a statistical sense, from a limited sample and using interpolation techniques in a multidimensional feature space, supervision being introduced through metric learning.…”

AI-driven prediction of SARS-CoV-2 variant binding trends from atomistic simulations

“…Indeed, artificial intelligence (AI) has recently emerged as a potential accelerator of MD simulations [30][31][32]. The majority of ML methods have focused on learning the force field [31], that is, the complex set of rules and parameters governing the interaction among atoms and thus between molecules.…”

“…The majority of ML methods have focused on learning the force field [31], that is, the complex set of rules and parameters governing the interaction among atoms and thus between molecules. Recently, an increasing amount of works devoted to the prediction of experimentally measurable quantities from MD has emerged [30,33,34]. In this paper we present a neural network-based method to predict molecular binding trends, that is, whether or not the binding affinity of a target molecular interaction is higher or lower than a reference interaction, and reduce the computational burden of the molecular simulations.…”

“…For MD simulations, the majority of ML methods have focused on learning the force field [41], that is, the complex set of rules and parameters governing the interaction among atoms and thus between molecules. Recently, an increasing amount of works devoted to the prediction of experimentally measurable quantities from MD has emerged [40, 45–47]. Specific to SARS-CoV-2, in a recent paper, MD simulations and ML methods are combined with FEP calculations to investigate the impact of individual residues on binding to ACE2 in SARS-CoV-2 compared to SARS-CoV [22].…”