Cluster-MLP: An Active Learning Genetic Algorithm Framework for Accelerated Discovery of Global Minimum Configurations of Pure and Alloyed Nanoclusters.

Structural characterization of nanoclusters is one of the major challenges in nanocluster modeling owing to the multitude of possible configurations of arrangement of cluster atoms. The genetic algorithm (GA), a class of evolutionary algorithms based on the principles of natural evolution, is a commonly employed search method for locating the global minimum configuration of nanoclusters. Although a GA search at the DFT level is required for the accurate description of a potential energy surface to arrive at the correct global minimum configuration of nanoclusters, computationally expensive DFT evaluation of the significantly larger number of cluster geometries limits its practicability. Recently, machine learning potentials (MLP) that are learned from DFT calculations gained significant attention as computationally cheap alternative options that provide DFT level accuracy. As the accuracy of the MLP predictions is dependent on the quality and quantity of the training DFT data, active learning (AL) strategies have gained significant momentum to bypass the need of large and representative training data. In this application note, we present Cluster-MLP, an on-the-fly active learning genetic algorithm framework that employs the Flare++ machine learning potential (MLP) for accelerating the GA search for global minima of pure and alloyed nanoclusters. We have used a modified version the Birmingham parallel genetic algorithm (BPGA) for the nanocluster GA search which is then incorporated into distributed evolutionary algorithms in Python (DEAP), an evolutionary computational framework for fast prototyping or technical experiments. We have shown that the incorporation of the AL framework in the BPGA significantly reduced the computationally expensive DFT calculations. Moreover, we have shown that both the AL-GA and DFT-GA predict the same global minima for all the clusters we tested.

Computational investigation of cobalt and copper bis (oxothiolene) complexes as an alternative for olefin purification.

Considering that olefins present a large volume feedstock, it is reasonable to expect that their purification is industrially critical. After the discovery of the nickel bis (dithiolene) complex Ni(S2C2(CF3)2)2 that exhibits electro-catalytic activity with olefins but tends to decompose by a competitive reaction route, related complexes have been explored experimentally and theoretically. In this paper, a computational examination is performed on differently charged cobalt and copper bis (oxothiolene) complexes [M (OSC2(CN)2)2] to test their potential applicability as the catalysts for olefin purification, using the simplest olefin, ethylene. Possible reaction pathways for ethylene addition on these complexes were explored, to determine whether some of these candidates can avoid the reaction route that leads to decomposition, which is distinctive from the nickel complex, and to form stable adducts that can subsequently release ethylene by reduction. Our calculations suggest that the neutral cobalt complex might be an alternative catalyst, because all its forms can bind ethylene to produce stable interligand adducts with moderate to low activation barriers, rather than to form intraligand adducts that lead to decomposition. The calculations also predict that these interligand adducts are capable of releasing ethylene upon reduction. In addition, it can produce the desired interligand adducts following two different reaction pathways, assigned as the direct and the indirect, with no need for anion species as co-catalysts, which is crucial for the nickel complex. Thus, the olefin purification process could be much simpler by using this catalyst.

Nickel Bis(diselenolene) as a Catalyst for Olefin Purification.

Nickel bis(dithiolene) reversibly binds olefins via a known interligand binding mechanism, but the complex has limited practical use, due to a competitive intraligand addition which results in decomposition. The present work examines an alternative nickel-based complex that eliminates the decomposition route. Specifically, we have examined the olefin binding processes of nickel bis(diselenolene) complexes using modern density functional theory. Both the inter- and intraligand adducts of the nickel bis(diselenolenes) are thermodynamically more stable than their dithiolene analogues. We have predicted that nickel bis(diselenolene) complexes do not decompose after the intraligand addition, and that the overall activation energies for the kinetically accessible products are quite small. In short, our computational work predicts that nickel bis(diselenolene) complexes are better electrocatalysts for olefin purification than the previous candidates, superior to the previously studied nickel bis(dithiolene) complexes.

A unified set of experimental organometallic data used to evaluate modern theoretical methods.

We applied a test set of ligand dissociation enthalpies derived entirely from a unified experimental approach to evaluate the efficacy of various methods for modeling organometallic chemistry. This differs from most benchmarking studies, as it is common to evaluate theoretical methods by using more computationally expensive calculations to provide the "target" values. With an aim of presenting the 'best suited functional/functionals' for calculations involving the metal-ligand bond dissociation enthalpies (BDEs) of organometallic complexes, we utilized a database of 30 experimental metal-ligand bond dissociation enthalpies, and tested for 101 density functionals and 2 ab initio methods, all with a large basis set. We find the most accurate functional is M06 with a mean unsigned error (MUE) of 1.6 kcal mol(-1), followed closely by M06L, ωB97XD, PW91PW91 and MPWB95 with MUEs of 1.7, 1.8, 2.0, and 2.1 kcal mol(-1) respectively. Other top performers are B3LYP-GD3, BLYP-GD3, PBEPBE, CAM-B3LYP-GD3, CAM-B3LYP-GD3BJ, B3LYP-GD3BJ and MN12L; all predict BDEs with MUEs in the range of 2.2 to 2.5 kcal mol(-1). Adding solvent corrections to the gas-phase BDE calculations for these top twelve functionals do not significantly change the MUE value. The well-known and widely used functional B3LYP shows very poor performance for this specific property. However, the empirical dispersion correction to the B3LYP functional has significantly improved its performance in predicting BDEs. It is also worth noting that several modern range-separated functionals predict the bond dissociation enthalpies with an error of 2-3 kcal mol(-1).

Mechanism of Ethylene Addition to Nickel Bis(oxothiolene) and Nickel Bis(dioxolene) Complexes.

The electrochemically reversible binding of olefins by nickel bis(dithiolene) has been extensively studied, both theoretically and computationally. To optimize a catalyst for this process, we have investigated all possible reaction pathways of ethylene addition onto the related complex nickel bis(dioxolene), and the two isomers (cis and trans) of nickel bis(oxothiolene). Modern DFT calculations predict that the nickel bis(dioxolene) complex has limited practical use due to high barriers to binding. However, each of the two isomers of the nickel bis(oxothiolene) complexes display enhanced properties versus the original nickel bis(dithiolene) complex. Specifically, in nickel bis(dithiolene), the intraligand binding of olefins leads to decomposition, whereas interligand binding is required for reversibility; the two nickel bis(oxothiolene) complexes have greater selectivity toward the formation of the desired interligand adducts. For the full reaction pathways, the new complexes' binding mechanisms are contrasted with the mechanism of the original catalyst.

Iron(0) mediated C-H activation of 1-hexyne: a mechanistic study using time-resolved infrared spectroscopy.

Photolysis of an iron tricarbonyl complex in the presence of 1-hexyne results in the activation of the terminal C-H bond to yield an iron-alkynyl species. The reaction proceeds through a single transition state with an activation enthalpy of 13.5 kcal mol(-1). The resulting molecule may have potential as a C-C bond formation reagent.

Aromatic interactions modulate the 5'-base selectivity of the DNA-binding autoantibody ED-10.

We present detailed computational analyses of the binding of four dinucleotides to a highly sequence-selective single-stranded DNA (ssDNA) binding antibody (ED-10) and selected point mutants. Anti-DNA antibodies are central to the pathogenesis of systemic lupus erythematosus (SLE), and a more complete understanding of the mode of binding of DNA and other ligands will be necessary to elucidate the role of anti-DNA antibodies in the kidney inflammation associated with SLE. Classical molecular mechanics based molecular dynamics simulations and density functional theory (DFT) computations were applied to pinpoint the origin of selectivity for the 5'-nucleotide. In particular, the strength of interactions between each nucleotide and the surrounding residues were computed using MMGBSA as well as DFT applied to a cluster model of the binding site. The results agree qualitatively with experimental binding free energies, and indicate that π-stacking, CH/π, NH/π, and hydrogen-bonding interactions all contribute to 5'-base selectivity in ED-10. Most importantly, the selectivity for dTdC over dAdC arises primarily from differences in the strength of π-stacking and XH/π interactions with the surrounding aromatic residues; hydrogen bonds play little role. These data suggest that a key Tyr residue, which is not present in other anti-DNA antibodies, plays a key role in the 5'-base selectivity, while we predict that the mutation of a single Trp residue can tune the selectivity for dTdC over dAdC.

Thermal and photochemical reactivity of manganese tricarbonyl and tetracarbonyl complexes with a bulky diazabutadiene ligand.

The manganese tricarbonyl complex fac-Mn(Br)(CO)3((i)Pr2Ph-DAB) (1) [(i)Pr2Ph-DAB = (N,N'-bis(2,6-di-isopropylphenyl)-1,4-diaza-1,3-butadiene)] was synthesized from the reaction of Mn(CO)5Br with the sterically encumbered DAB ligand. Compound 1 exhibits rapid CO release under low power visible light irradiation (560 nm) suggesting its possible use as a photoCORM. The reaction of compound 1 with TlPF6 in the dark afforded the manganese(I) tetracarbonyl complex, [Mn(CO)4((i)Pr2Ph-DAB)][PF6] (2). While 2 is comparatively more stable than 1 in light, it demonstrates high thermal reactivity such that dissolution in CH3CN or THF at room temperature results in rapid CO loss and formation of the respective solvate complexes. This unusual reactivity is due to the large steric profile of the DAB ligand which results in a weak Mn-CO binding interaction.

Insights into the activity and specificity of Trypanosoma cruzi trans-sialidase from molecular dynamics simulations.

Trypanosoma cruzitrans-sialidase (TcTS), which catalyzes the transfer or hydrolysis of terminal sialic acid residues, is crucial to the development and proliferation of the T. cruzi parasite and thus has emerged as a potential drug target for the treatment of Chagas disease. We here probe the origin of the observed preference for the transfer reaction over hydrolysis where the substrate for TcTS is the natural sialyl donor (represented in this work by sialyllactose). Thus, acceptor lactose preferentially attacks the sialyl-enyzme intermediate rather than water. We compare this with the weaker preference for such transfer shown by a synthetic donor substrate, 4-methylumbelliferyl α-d-acetylneuraminide. For this reason, we conducted molecular dynamics simulations of TcTS following its sialylation by the substrate to examine the behavior of the asialyl leaving group by the protein. These simulations indicate that, where lactose is released, this leaving group samples well-defined interactions in the acceptor site, some of which are mediated by localized water molecules; also, the extent of the opening of the acceptor site to solvent is reduced as compared with those of unliganded forms of TcTS. However, where there is release of 4-methylumbelliferone, this leaving group explores a range of transient poses; surrounding active site water is also more disordered. The acceptor site explores more open conformations, similar to the case in which the 4-methylumbelliferone is absent. Thus, the predicted solvent accessibility of sialylated TcTS is increased when 4-methylumbelliferyl α-d-acetylneuraminide is the substrate compared to sialyllactose; this in turn is likely to contribute to a greater propensity for hydrolysis of the covalent intermediate. These computational simulations, which suggest that protein flexibility has a role in the transferase/sialidase activity of TcTS, have the potential to aid in the design of anti-Chagas inhibitors effective against this neglected tropical disease.

Broad Transferability of Substituent Effects in π-Stacking Interactions Provides New Insights into Their Origin.

Substituent effects in model stacked homodimers and heterodimers of benzene, borazine, and 1,3,5-triazine have been examined computationally. We show that substituent effects in these dimers are strongly dependent on the identity of the unsubstituted ring, yet are independent of the ring bearing the substituent. This supports the local, direct interaction model [J. Am. Chem. Soc. 2011, 133, 10262], which maintains that substituent effects in π-stacking interactions are dominated by through-space interactions of the substituents with the proximal vertex of the unsubstituted ring. In addition to dimers in which the unsubstituted ring is held constant, substituent effects are correlated in many other stacked dimers, including those in which neither the substituted nor unsubstituted rings are conserved. Whether substituent effects in a pair of dimers will be correlated is shown to hinge on the electrostatic components of the interaction energies, and the correlations are explained in terms of the interaction of the local dipole moments associated with the substituents and the electric fields of the unsubstituted rings. Overall, substituent effects are similar in two stacked dimers as long as the electric fields above the unsubstituted rings are similar, providing a more sound physical justification for the local, direct interaction model.

Physical Nature of Substituent Effects in XH/π Interactions.

XH/π interactions (e.g.: CH/π, OH/π, etc.) are ubiquitous in chemical and biochemical contexts. Although there have been many studies of substituent effects in XH/π interactions, there have been only limited systematic studies covering a broad range of substituents. We provide a comprehensive and systematic study aimed at unraveling the nature of aryl substituent effects on model BH/π, CH/π, NH/π, OH/π, and F/π interactions (e.g.: BH3···C6H5Y, CH4···C6H5Y, etc.) based on estimated CCSD(T)/aug-cc-pVTZ interaction energies as well as symmetry-adapted perturbation theory (SAPT) results. We show that the impact of substituents on XH/π interactions depends strongly on the identity of the XH group, and the strength of these effects increases with increasing polarization of the XH bond. Overall, the results are in accord with previous work and follow expected trends from basic physical principles. That is, electrostatic effects dominate the substituent effects for the polar XH/π interactions (NH/π, OH/π, and FH/π), while dispersion effects are more important for the nonpolar BH/π and CH/π interactions. The electrostatic component of these interactions is shown to correlate well with Hammett constants (σm), while accounting for the dispersion component requires consideration of molar refractivities (MR) and interaction distances concurrently. The correlation of the dispersion component of these interactions with MR values alone is rather weak.

Substituent effects on non-covalent interactions with aromatic rings: insights from computational chemistry.

Non-covalent interactions with aromatic rings pervade modern chemical research. The strength and orientation of these interactions can be tuned and controlled through substituent effects. Computational studies of model complexes have provided a detailed understanding of the origin and nature of these substituent effects, and pinpointed flaws in entrenched models of these interactions in the literature. Here, we provide a brief review of efforts over the last decade to unravel the origin of substituent effects in π-stacking, XH/π, and ion/π interactions through detailed computational studies. We highlight recent progress that has been made, while also uncovering areas where future studies are warranted.

An evaluation of the GLYCAM06 and MM3 force fields, and the PM3-D* molecular orbital method for modelling prototype carbohydrate-aromatic interactions.

The structures and interaction energies of 21 binary complexes of fucose and glucose with toluene, 3-methylindole or p-hydroxytoluene, evaluated at the DFT-D level, are used to judge the accuracy of the GLYCAM06 and MM3 force fields, and the PM3-D* molecular orbital method for modelling carbohydrate-arene interactions. The accuracy of the DFT-D method is substantiated by comparison with high level CCSD(T) calculations on a small number of representative complexes. It is found that a correct description of the intermolecular dispersive interactions is essential. Both the PM3-D* method and the GYLCAM06 force field yield interaction energies within 1 kcal mol⁻¹ of the DFT-D values, whilst those from the MM3 force field are in error by more than 2 kcal mol⁻¹.

The effects of perfluorination on carbohydrate-pi interactions: computational studies of the interaction of benzene and hexafluorobenzene with fucose and cyclodextrin.

The effect of benzene fluorination on C-H...pi interactions is studied using a number of computational methods applied to a range of intermolecular complexes. High level wavefunction methods (CCSD(T)) predict a slightly greater interaction energy for complexes of benzene with methane or fucose, compared to corresponding complexes involving hexafluorobenzene. A number of more approximate treatments, DFT with the M06-2X functional, PM3-D* and MM methods, give interaction energies within 1 kcal mol(-1) of the high level values, and also correctly predict that the interaction energy is slightly greater for benzene compared to hexafluorobenzene. However, the DFT-D model used here predicts that the interaction energy is slightly greater for hexafluorobenzene. Molecular dynamics simulations, employing the GLYCAM-06 force field, validated here, are used to model the complexes of benzene and hexafluorobenzene with beta-cyclodextrin in aqueous solution. We predict the binding free energies of the complexes to be within 0.5 kcal mol(-1), and suggest that the different chemical shifts of the H5 protons observed in the two complexes arise from their slightly different structures, rather than from different binding energies.

Modelling the binding of HIV-reverse transcriptase and nevirapine: an assessment of quantum mechanical and force field approaches and predictions of the effect of mutations on binding.

The importance of the intermolecular interactions which contribute to the binding of HIV-1 RT with the NNRTI inhibitor, nevirapine (NVP), has been studied using quantum mechanical and molecular simulation methods. A range of computational methods, including density functional theory with empirical dispersion corrections, have been employed and show that although pi-pi stacking interactions are important, the combined effect of a number of C-H/pi interactions provides a significant contribution to the binding. The AMBER empirical force-field has been shown to be particularly effective to describe the interactions in this case; MM-GBSA free-energy methods were subsequently used to explore the effects on binding with several known mutations of HIV-1 RT. The relative affinities from the mutation simulations are shown to be in good agreement with experimental data allowing the causes of the binding changes to be discussed.

Carbohydrate-aromatic pi interactions: a test of density functionals and the DFT-D method.

The performance of a number of computational approaches based upon density functional theory (DFT) for the accurate description of carbohydrate-pi interactions is described. A database containing interaction energies of a small number of representative complexes, computed at a high ab initio level, is described, and is used to judge 18 different density functionals including the M05 and M06 families as well as the DFT method augmented with empirical dispersive corrections (DFT-D). The DFT-D method and the M06 functionals are found to perform particularly well, whilst traditional functionals such as B3LYP perform poorly. The interaction energies for 23 sugar-aromatic complexes calculated by the DFT-D method are compared with the values from the 18 functionals. Again, the M06 class of functional is found to be superior.

Carbohydrate-protein recognition probed by density functional theory and ab initio calculations including dispersive interactions.

Carbohydrate-protein recognition has been studied by electronic structure calculations of complexes of fucose and glucose with toluene, p-hydroxytoluene and 3-methylindole, the latter aromatic molecules being analogues of phenylalanine, tyrosine and tryptophan, respectively. We use mainly a density functional theory model with empirical corrections for the dispersion interactions (DFT-D), this method being validated by comparison with a limited number of high level ab initio calculations. We have calculated both binding energies of the complexes as well as their harmonic vibrational frequencies and proton NMR chemical shifts. We find a range of minimum energy structures in which the aromatic group can bind to either of the two faces of the carbohydrate, the binding being dominated by a combination of OH-pi and CH-pi dispersive interactions. For the fucose-toluene and alpha-methyl glucose-toluene complexes, the most stable structures involve OH-pi interactions, which are reflected in a red shift of the corresponding O-H stretching frequency, in good quantitative agreement with experimental data. For those structures where CH-pi interactions are found we predict a corresponding blue shift in the C-H frequency, which parallels the predicted proton NMR shift. We find that the interactions involving 3-methylindole are somewhat greater than those for toluene and p-hydroxytoluene.

The interaction of carbohydrates and amino acids with aromatic systems studied by density functional and semi-empirical molecular orbital calculations with dispersion corrections.

Density functional theory (DFT-D) and semi-empirical (PM3-D) methods having an added dispersion correction have been used to study stabilising carbohydrate-aromatic and amino acid-aromatic interactions. The interaction energy for three simple sugars in different conformations with benzene, all give interaction energies close to 5 kcal mol(-1). Our original parameterization of PM3 (PM3-D) seriously overestimates this value, and has prompted a reparametrization which includes a modified core-core interaction term. With two additional parameters, the carbohydrate complexes, as well as the S22 data set, are well reproduced. The new PM3 scheme (PM3-D*) is found to describe the peptide bond-aromatic ring interactions accurately and, together with the DFT-D method, it is used to investigate the interaction of six amino acids with pyrene. Whilst the peptide backbone can adopt both stacked and T-shaped structures in the complexes with similar interaction energies, there is a preference for the unsaturated ring to adopt a stacked structure. Thus, peptides in which the latter interactions are maximised are likely to be the most effective for the functionalisation of carbon nanotubes.