Identifying and embedding transferability in data-driven representations of chemical space.

Transferability, especially in the context of model generalization, is a paradigm of all scientific disciplines. However, the rapid advancement of machine learned model development threatens this paradigm, as it can be difficult to understand how transferability is embedded (or missed) in complex models developed using large training data sets. Two related open problems are how to identify, without relying on human intuition, what makes training data transferable; and how to embed transferability into training data. To solve both problems for ab initio chemical modelling, an indispensable tool in everyday chemistry research, we introduce a transferability assessment tool (TAT) and demonstrate it on a controllable data-driven model for developing density functional approximations (DFAs). We reveal that human intuition in the curation of training data introduces chemical biases that can hamper the transferability of data-driven DFAs. We use our TAT to motivate three transferability principles; one of which introduces the key concept of transferable diversity. Finally, we propose data curation strategies for general-purpose machine learning models in chemistry that identify and embed the transferability principles.

Sorting drug conformers in enzyme active sites: the XTB way.

In the present study, we have used the MEI196 set of interaction energies to investigate low-cost computational chemistry approaches for the calculation of binding between a molecule and its environment. Density functional theory (DFT) methods, when used with the vDZP basis set, yield good agreement with the reference energies. On the other hand, semi-empirical methods are less accurate as expected. By examining different groups of systems within MEI196 that contain species of a similar nature, we find that chemical similarity leads to cancellation of errors in the calculation of relative binding energies. Importantly, the semi-empirical method GFN1-xTB (XTB1) yields reasonable results for this purpose. We have thus further assessed the performance of XTB1 for calculating relative energies of docking poses of substrates in enzyme active sites represented by cluster models or within the ONIOM protocol. The results support the observations on error cancellation. This paves the way for the use of XTB1 in parts of large-scale virtual screening workflows to accelerate the drug discovery process.

Limiting factors in the accuracy of DFT calculation for redox potentials.

In the present study, we have investigated factors affecting the accuracy of computational chemistry calculation of redox potentials, namely the gas-phase ionization energy (IE) and electron affinity (EA), and the continuum solvation effect. In general, double-hybrid density functional theory methods yield IEs and EAs that are on average within ~0.1 eV of our high-level W3X-L benchmark, with the best performing method being DSD-BLYP/ma-def2-QZVPP. For lower-cost methods, the average errors are ~0.2-0.3 eV, with ωB97X-3c being the most accurate (~0.15 eV). For the solvation component, essentially all methods have an average error of ~0.3 eV, which shows the limitation of the continuum solvation model. Curiously, the directly calculated redox potentials show errors of ~0.3 eV for all methods. These errors are notably smaller than what can be expected from error propagation with the two components (IE and EA, and solvation effect). Such a discrepancy can be attributed to the cancellation of errors, with the lowest-cost GFN2-xTB method benefiting the most, and the most accurate ωB97X-3c method benefiting the least. For organometallic species, the redox potentials show large deviations exceeding ~0.5 eV even for DSD-BLYP. The large errors are attributed to those for the gas-phase IEs and EAs, which represents a major barrier to the accurate calculation of redox potentials for such systems.

Atomic-resolution structure analysis inside an adaptable porous framework.

We introduce a versatile metal-organic framework (MOF) for encapsulation and immobilization of various guests using highly ordered internal water network. The unique water-mediated entrapment mechanism is applied for structural elucidation of 14 bioactive compounds, including 3 natural product intermediates whose 3D structures are clarified. The single-crystal X-ray diffraction analysis reveals that incorporated guests are surrounded by hydrogen-bonded water networks inside the pores, which uniquely adapt to each molecule, providing clearly defined crystallographic sites. The calculations of host-solvent-guest structures show that the guests are primarily interacting with the MOF through weak dispersion forces. In contrast, the coordination and hydrogen bonds contribute less to the total stabilization energy, however, they provide highly directional point interactions, which help align the guests inside the pore.

Methide Affinity Scale: Key Thermodynamic Data Underpinning Catalysis, Organic Synthesis, and Organometallic and Polymer Chemistry.

Methide transfer reactions play important roles in many areas of chemistry, including the Grignard reaction, in the transmetalation steps of metal-catalyzed cross-coupling reactions, and in the generation of cationic metal polymerization catalysts. Methide affinities (MAs) are the key thermodynamic quantity that underpin such reactions, and yet comprehensive methide affinity scales are poorly developed. Here, B3LYP-D3BJ/def2-TZVP calculations are used to calculate the energy changes (MAs) for cations (MeZ → Z+ + Me-), neutrals (MeY- → Y + Me-), and anions (MeX2- → X- + Me-) derived from permethyl species of all group s and p elements. The MAs range from 2525.8 for the singlet cation F+ to -820.4 kJ/mol for the tetramethylborate anion, Me4B-. The cations show the clearest trends: the MAs in all cases decrease going down the group, while moving across a period, the MAs increase from group 1 to group 2 and then decrease for group 3, remaining about the same or with a modest increase moving to group 4, and then continue to increase across a period to a maximum for the halogens (group 17). The anions and dianions are sensitive to hypervalency; those elements that cannot expand the octet have very unfavorable MAs (e.g., MA of Me4C requires the formation of Me5C- and of Me4B- requires the formation of Me5B2-). To address whether the anion MeY- and dianion MeZ2- are stable, the vertical detachment energies of the anions and dianions were calculated. All of the anions are thermodynamically stable with respect to electron loss, except for Me4N-, while the dianions are all thermodynamically unstable with respect to electron loss. The kinetic stability of the dianions with respect to methide and electron loss was also evaluated for the lowest MAs. The only dianions that might be kinetically stable and observable in the gas phase are Me4Ca2-, Me4Sr2-, and Me4Ba2-. The dianion CF3CaF32- is predicted to be both thermodynamically and kinetically stable in the gas phase.

The verification of delta SCF and Slater's transition state theory for the calculation of core ionization energy.

The core ionization energies of second- and third-period elements of the molecules C2 H5 NO2 , SiF4 , Si(CH3 )4 , PF3 , POF3 , PSF3 , CS2 , OCS, SO2 , SO2 F2 , CH3 Cl, CFCl3 , SF5 Cl, and Cl3 PS are calculated by using Hartree-Fock (HF), and Kohn-Sham (KS) with BH&HLYP, B3LYP, and LC-BOP functionals. We used ΔSCF, Slater's transition state (STS), and two previously proposed shifted STS (1) and shifted STS (2) methods, which have been developed. The errors of ΔSCF and STS come mainly from the self-interaction errors (SIE) and can be corrected with a shifting scheme. In this study, we used the shifting parameters determined for each atom. The shifted STS (1) reproduces ΔSCF almost perfectly with mean absolute deviations (MAD) of 0.02 eV. While ΔSCF and STS vary significantly depending on the functional used, the variation of shifted STS (2) is small, and all shifted STS (2) values are close to the observed ones. The deviations of the shifted STS (2) from the experiment are 0.24 eV (BH&HLYP), 0.19 eV (B3LYP), and 0.23 eV (LC-BOP). These results further support the use of shifted STS methods for predicting the core ionization energies.

DAPD Set of Pd-Containing Diatomic Molecules: Accurate Molecular Properties and the Great Lengths to Obtain Them.

In the present study, we obtained reliable bond energy, bond length, and zero-point vibrational frequency for a set of diatomic Pd species (the DAPD set). It includes PdH, Pd2, and PdX (X = B, C, N, O, F, Al, Si, P, S, and Cl). Our highest-level protocol (W4X-L) represents scalar and spin-orbit relativistic, valence- and inner-valence correlated, extrapolated CCSDTQ(5) energy. The DAPD set of molecules is challenging for computational chemistry methods in different manners; for Pd2, the spin-orbit contribution to the bond energy is fairly large, whereas for PdC and PdSi, the post-CCSD(T) correlation components are considerable. The diverse range of requirements represents a significant challenge for lower-level methods. While density functional theory (DFT) methods generally yield good agreements for bond lengths and vibrational frequencies, large deviations are found for bond energies. In general, hybrid DFT methods are more accurate than nonhybrid functionals, but the agreement in individual cases varies. This illustrates the critical role that new high-quality reference data would play in the continual development of lower-cost methods.

Dynamic and static control of the off-target interactions of antisense oligonucleotides using toehold chemistry.

Off-target interactions between antisense oligonucleotides (ASOs) with state-of-the-art modifications and biological components still pose clinical safety liabilities. To mitigate a broad spectrum of off-target interactions and enhance the safety profile of ASO drugs, we here devise a nanoarchitecture named BRace On a THERapeutic aSo (BROTHERS or BRO), which is composed of a standard gapmer ASO paired with a partially complementary peptide nucleic acid (PNA) strand. We show that these non-canonical ASO/PNA hybrids have reduced non-specific protein-binding capacity. The optimization of the structural and thermodynamic characteristics of this duplex system enables the operation of an in vivo toehold-mediated strand displacement (TMSD) reaction, effectively reducing hybridization with RNA off-targets. The optimized BROs dramatically mitigate hepatotoxicity while maintaining the on-target knockdown activity of their parent ASOs in vivo. This technique not only introduces a BRO class of drugs that could have a transformative impact on the extrahepatic delivery of ASOs, but can also help uncover the toxicity mechanism of ASOs.

Interfacial alloying between lead halide perovskite crystals and hybrid glasses.

The stellar optoelectronic properties of metal halide perovskites provide enormous promise for next-generation optical devices with excellent conversion efficiencies and lower manufacturing costs. However, there is a long-standing ambiguity as to whether the perovskite surface/interface (e.g. structure, charge transfer or source of off-target recombination) or bulk properties are the more determining factor in device performance. Here we fabricate an array of CsPbI3 crystal and hybrid glass composites by sintering and globally visualise the property-performance landscape. Our findings reveal that the interface is the primary determinant of the crystal phases, optoelectronic quality, and stability of CsPbI3. In particular, the presence of a diffusion "alloying" layer is discovered to be critical for passivating surface traps, and beneficially altering the energy landscape of crystal phases. However, high-temperature sintering results in the promotion of a non-stoichiometric perovskite and excess traps at the interface, despite the short-range structure of halide is retained within the alloying layer. By shedding light on functional hetero-interfaces, our research offers the key factors for engineering high-performance perovskite devices.

Simple Composite Approach to Efficiently Estimate Basis Set Limit CCSD(T) Harmonic Frequencies and Reaction Thermochemistry.

We introduce a simple strategy that combines the G3(MP2) composite method and explicitly correlated coupled cluster CCSD(T)-F12 method to efficiently estimate complete basis set CCSD(T) molecular geometries and harmonic vibrational frequencies at the cost of a double-ζ basis set calculation. Based on a large test set of 61 neutral, ionic, and open-shell molecules, and additionally 31 molecules in the HFREQ2014 data set, we demonstrate that this composite strategy has an average accuracy of 2 cm-1 or better relative to complete basis set CCSD(T) values. Using this approach, we estimated 696 CCSD(T)/CBS reaction energies of small to medium-sized systems containing up to 6 heavy atoms and confirmed existing approximations that use small basis set density functional theory methods [e.g., M06-2X/6-31+G(d)] to calculate thermal contributions to reaction enthalpies and Gibbs free energies that are accurate to within 0.2 kcal mol-1 on average.

Core-Level 2s and 2p Binding Energies of Third-Period Elements (P, S, and Cl) Calculated by Hartree-Fock and Kohn-Sham ΔSCF Theory.

In the present study, we investigate the use of the ΔSCF method and Slater's transition state (STS) theory to calculate the binding energies of the 2s and 2p electrons of third-period elements (P, S, and Cl). Both the Hartree-Fock (HF) and Kohn-Sham (KS) approximations are examined. The STS approximation performs well in reproducing the ΔSCF values. However, for the ΔSCF method itself, while the binding energy of the 2p electrons is accurately predicted, the results for 2s are fairly sensitive to the functional, exhibiting significant variations due to self-interaction errors (SIE). Nonetheless, the variations in chemical shifts between different species remain relatively small, and the values agree with experiments due to the cancellation of SIE. A notable observation is that the chemical shifts of the 2s and 2p electrons are similar, indicating a perturbation caused by the valence electrons. The error in the absolute binding energy of KS ΔSCF against the experiment is nearly constant for the same element in different molecules, and it depends largely on the functional owing to SIE. A shifting scheme previously developed can be employed to reproduce the experimental 2s and 2p binding energies, with dependence on the functional and atom but not on the molecule even for 2s KS ΔSCF binding energies. Upon obtaining the corrected binding energies, we find that the gap between 2s and 2p binding energy is nearly independent of chemical environment for a given element: 57.5, 63.9, and 70.9 eV for the elements P, S, and Cl, respectively.

Compilation of Ionic Clusters with the Rock Salt Structure: Accurate Benchmark Thermochemical Data, Assessment of Quantum Chemistry Methods, and the Convergence Behavior of Lattice Energies.

In the present study, computational quantum chemistry is used to obtain lattice energies (LEs) for a range of ionic clusters with the NaCl structure. Specifically, the compounds include NaF, NaCl, MgO, MgS, KF, CaO, and CaS clusters, (MX)n, with n = 1, 2, 4, 6, 8, 12, 16, 24, 32, 40, 50, 60, 75, 90, and 108. The highest-level W2 and W1X-2 methods are applied to the small clusters with n = 1 to 8 (the MX35 data set). The assessment with MX35 shows that, for the calculation of geometries and vibrational frequencies, the PBE0-D3(BJ) and PBE-D3(BJ) DFT methods are reasonable, but the calculation of atomization energies is more challenging. This is a result of different systematic deviations for clusters of different species. Thus, species-specific adjustments are applied for larger clusters, which are calculated with the DuT-D3 double-hybrid DFT method, the MN15 DFT method, and the PM7 semi-empirical method. They yield smoothly converging LEs to the bulk values. It is also found that, for the alkali-metal species, the LEs for a single molecule are ∼70% of the bulk values, while for the alkali-earth species, they are ∼80%. This has enabled a straightforward means to the first-principles estimation of LEs for similarly structured ionic compounds.

Optimal Small Basis Set and Geometric Counterpoise Correction for DFT Computations.

In the present study, we have examined the performance of the various small basis sets and their geometric counterpoise (gCP) corrections for DFT computations. The original gCP correction scheme includes four adjustable parameters tailored for each method and basis set, but we find that the use of a single scaling parameter also yields fair results. We term this simplified scheme unity-gCP, which can be straightforwardly applied for devising a reasonable correction for an arbitrary basis set. With the use of unity-gCP, we have examined a systematic set of medium-sized basis sets, and we find 6-31+G(2d) to be the optimal balance between accuracy and computational efficiency. On the other hand, less balanced basis sets, even larger ones, can show significantly worse accuracy; the inclusion of gCP may even lead to severe overcorrections. Thus, sufficient validations would be imperative before the general application of gCP for a particular basis set. For 6-31+G(2d), a welcoming finding is that its gCP has small magnitudes, and thus, it also yields adequate results without gCP corrections. This observation echoes that for the ωB97X-3c method, which uses an optimized double-ζ basis set (vDZP) without the inclusion of gCP. In an attempt to improve vDZP by mimicking the somewhat better-performing 6-31+G(2d), we partially decontract the outer functions of vDZP. The resulting basis set, which we termed vDZ+(2d), generally yields improved results. Overall, the vDZP and the new vDZ+(2d) basis sets pave a way for obtaining reasonable results more efficiently for a wide range of systems than the practice of using a triple- or quadruple-ζ basis set in DFT calculations.

Computational insights into the singlet-triplet energy gaps, ionization energies, and electron affinities for a diverse set of 812 small fullerenes (C20-C50).

In the present study, we have investigated the energy differences between the lowest-energy singlet and triplet states of a large set of small fullerenes with density functional theory (DFT), and the related quantities of ionization energy (IE) and electron affinity (EA). The DFT methods generally show consistent qualitative observations. For the full set of 812 fullerene isomers, ∼80-90% have a singlet ground state, with the rest being ground-state triplets; some of them may complement existing singlet-fission materials to improve the efficiency for light harvesting. The triplet-singlet energy difference correlates well with the IE-EA differences, which are indicators for charge-transfer capabilities. We have surveyed larger fullerenes in search of candidates with superior charge-transfer properties, with the results suggesting that optimally shaped medium-sized fullerenes may be the most promising.

The core ionization energies calculated by delta SCF and Slater's transition state theory.

The core ionization energies of the second-period and third-period elements are studied by ΔSCF and Slater's transition state (STS) theory by using Hartree-Fock (HF) and Kohn-Sham (KS) approximations. Electron correlation increases the estimated core ionization energies, while the self-interaction error (SIE) decreases them, especially for the third-period elements and is a more significant factor. As a result, while HF lacks electron correlation, it is free of SIE and reasonably predicts the core ionization energies. The core ionization energies calculated by HF STS are very close to those calculated by HF ΔSCF, showing that STS reasonably describes the relaxation of the core hole. The core ionization energies calculated by KS are particularly sensitive to the SIE of the functional used, with functionals having less SIE yielding more accurate ΔSCF core ionization energies. Consequently, BH&HLYP gives better results than B3LYP and LC-BOP since BH&HLYP is the hybrid functional with high proportion of the exact HF exchange. Although the core ionization energies are underestimated by ΔSCF due to SIE, STS gives larger core ionization energies than ΔSCF due to a concave behavior of the error curves of STS, which is also related to SIE. The mean absolute deviations of STS relative to ΔSCF, and relative to the experiment, are almost constant regardless of the nuclei among the element in the second period, and likewise among those in the third period. The systematic nature suggests that shifting the STS core ionization energies may be useful. We propose the shifted STS (1) for reproducing ΔSCF values, and the shifted STS (2) to reproduce the observed ones for KS calculations. Both schemes work quite well. The calculated results of KS ΔSCF and STS vary depending on the functional. However, the variation of each species' shifted STS (2) is very small, and all shifted STS (2) values are close to the observed ones. As the shifted STS require only one SCF calculation, they are simple and practical for predicting the core ionization energies.

Generation and reaction of alanyl radicals in open flasks.

The generation and Giese-type reaction of alanyl radicals under metal-free reaction conditions is described. The procedure is operationally simple, occurring at ambient temperature in an open reaction vessel, and requiring short reaction times (≤5 min). The reaction occurs without epimerization and provides ready access to non-proteinogenic amino acids and peptides. Importantly, the process is tolerant of light absorbing groups including commonly used fluorescent tags.

Experimental-theoretical study of laccase as a detoxifier of aflatoxins.

We investigate laccase-mediated detoxification of aflatoxins, fungal carcinogenic food contaminants. Our experimental comparison between two aflatoxins with similar structures (AFB1 and AFG2) shows significant differences in laccase-mediated detoxification. A multi-scale modeling approach (Docking, Molecular Dynamics, and Density Functional Theory) identifies the highly substrate-specific changes required to improve laccase detoxifying performance. We employ a large-scale density functional theory-based approach, involving more than 7000 atoms, to identify the amino acid residues that determine the affinity of laccase for aflatoxins. From this study we conclude: (1) AFB1 is more challenging to degrade, to the point of complete degradation stalling; (2) AFG2 is easier to degrade by laccase due to its lack of side products and favorable binding dynamics; and (3) ample opportunities to optimize laccase for aflatoxin degradation exist, especially via mutations leading to π-π stacking. This study identifies a way to optimize laccase for aflatoxin bioremediation and, more generally, contributes to the research efforts aimed at rational enzyme optimization.

Chiral Bifunctional Selenide Catalysts for Asymmetric Iodolactonizations.

1,1'-Bi-2-naphthol (BINOL)-derived chiral bifunctional sulfide and selenide catalysts that possess a hydroxy group are known to be effective catalysts for enantioselective bromolactonizations. When applied to asymmetric iodolactonizations of 4-pentenoic acids, these catalysts yield chiral γ-butyrolactone products that are important compounds in medicinal chemistry. Although chiral bifunctional selenides have shown good catalytic performances in enantioselective iodolactonizations, reactions with BINOL-derived chiral sulfide catalysts unexpectedly gave iodolactonization products in nearly racemic forms. The roles of chalcogenide moieties and hydroxy groups on bifunctional catalysts were investigated, and the importance of both a selenide moiety and a hydroxy group on chiral bifunctional selenide catalysts to achieve enantioselective iodolactonizations was clarified. An optimized chiral bifunctional selenide catalyst was applied to the asymmetric synthesis of chiral γ-butyrolactones and phthalides. Furthermore, the utility of chiral bifunctional selenides was also demonstrated in the catalytic enantioselective desymmetrizing iodolactonization of α,α-diallyl carboxylic acids.

High-Level Quantum Chemistry Reference Heats of Formation for a Large Set of C, H, N, and O Species in the NIST Chemistry Webbook and the Identification and Validation of Reliable Protocols for Their Rapid Computation.

A recent study has examined the accuracy of NIST heats of formation for a set of C, H, and O-containing species with a proposed low-cost quantum chemistry approach. In the present study, we have used high-level methods such as W1X-2 to obtain these data more rigorously, which we have then used to assess the NIST and the previously computed values. We find that many of these NIST data that are as suggested to be unreliable by the previous study are indeed inconsistent with our high-level reference values. However, we also find substantial deviations for the previously computed values from our benchmark. Thus, we have assessed the performance of alternative low-cost methods. In our assessment, we have additionally examined C, H, N, and O-containing species for which heats of formation are available from the NIST database. We find the ωB97M-V/ma-def2-TZVP, DSD-PBEP86/ma-def2-TZVP, and CCSD(T)-F12b/aug'-cc-pVDZ methods to be adequate for obtaining heats of formation with the atomization approach, once their atomic energies are optimized with our benchmark. Notably, the low-cost ωB97M-V method yields values that agree to be within 10 kJ mol-1 for more than 90% of the (∼1500) species. A higher 20 kJ mol-1 threshold captures 98% of the data. The outlier species typically contain many electron-withdrawing (nitro) groups. In these cases, the use of isodesmic-type reactions rather than the atomization approach is more reliable. Our assessment has also identified significant outliers from the NIST database, for which experimental re-determination of the heats of formation would be desirable.