Reagent prediction with a molecular transformer improves reaction data quality.

Automated synthesis planning is key for efficient generative chemistry. Since reactions of given reactants may yield different products depending on conditions such as the chemical context imposed by specific reagents, computer-aided synthesis planning should benefit from recommendations of reaction conditions. Traditional synthesis planning software, however, typically proposes reactions without specifying such conditions, relying on human organic chemists who know the conditions to carry out suggested reactions. In particular, reagent prediction for arbitrary reactions, a crucial aspect of condition recommendation, has been largely overlooked in cheminformatics until recently. Here we employ the Molecular Transformer, a state-of-the-art model for reaction prediction and single-step retrosynthesis, to tackle this problem. We train the model on the US patents dataset (USPTO) and test it on Reaxys to demonstrate its out-of-distribution generalization capabilities. Our reagent prediction model also improves the quality of product prediction: the Molecular Transformer is able to substitute the reagents in the noisy USPTO data with reagents that enable product prediction models to outperform those trained on plain USPTO. This makes it possible to improve upon the state-of-the-art in reaction product prediction on the USPTO MIT benchmark.

Exploring Chemical Reaction Space with Reaction Difference Fingerprints and Parametric t-SNE.

Humans prefer visual representations for the analysis of large databases. In this work, we suggest a method for the visualization of the chemical reaction space. Our technique uses the t-SNE approach that is parameterized using a deep neural network (parametric t-SNE). We demonstrated that the parametric t-SNE combined with reaction difference fingerprints could provide a tool for the projection of chemical reactions on a low-dimensional manifold for easy exploration of reaction space. We showed that the global reaction landscape projected on a 2D plane corresponds well with the already known reaction types. The application of a pretrained parametric t-SNE model to new reactions allows chemists to study these reactions in a global reaction space. We validated the feasibility of this approach for two commercial drugs, darunavir and montelukast. We believe that our method can help to explore reaction space and will inspire chemists to find new reactions and synthetic ways.

Macrocycle- and metal-centered reduction of metal tetraphenylporphyrins where the metal is copper(II), nickel(II) and iron(II).

The reduction of metal(II) tetraphenylporphyrins, where metal(II) is copper, nickel or iron, has been performed in toluene solution in the presence of a cryptand. Cesium anthracenide was used as a reductant. Crystalline salts {cryptand(Cs+)}2{CuII(TPP4-)}2- (1) and {cryptand(Cs+)}{NiI(TPP2-)}-·C6H5CH3 (2) have been obtained. The two-electron reduction of {CuII(TPP2-)}0 is centered on the macrocycle allowing one to study for the first time the structure and properties of the TPP4- tetraanions in the solid state. Tetraanions have a diamagnetic state and show essential C-Cmeso bond alternation. New bands attributed to TPP4- appear at 670, 770 and 870 nm. Unpaired S = 1/2 spin is localized on CuII. The one-electron reduction of {NiII(TPP2-)}0 centered on nickel provides the formation of {NiI(TPP2-)}- with unpaired S = 1/2 spin localized on NiI at 100(2) K. The effective magnetic moment of 2 is 1.68µB at 120 K and a broad asymmetric EPR signal characteristic of NiI is observed for 2 and also for (Bu3MeP+){NiI(TPP2-)}-·C6H5CH3 (3) in the 4.2-120 K range. Since dianionic TPP2- macrocycles are present at 100(2) K, no alternation of C-Cmeso bonds is observed in 2. The excited quartet S = 3/2 state according to the calculations is positioned close to the ground S = 1/2 state. In the excited state, charge transfer from NiI to the macrocycle takes place resulting in the formation of NiII with S = 1 and TPP˙3- with S = 1/2 in the {NiII(TPP˙3-)}- anions. Therefore, the increase in the magnetic moment of 2 above 150 K is attributed to the population of the excited quartet state with a gap of 750 K. Salt 2 is EPR silent above 150 K and manifests absorption bands characteristic of TPP˙3- at RT. The reduction of NiII(TPP2-) and FeII(TPP2-) by cesium anthracenide in the presence of Bu3MeP+ yields crystals of 3 and (Bu3MeP+){FeI(TPP2-)}-·C6H5CH3 (4) whose crystal structures and optical properties are also presented. DFT calculations have been carried out for {MII(TPP2-)} (M = Cu, Ni and Fe) and their anions to interpret the experimental results obtained for 1-4.

Effect of reduction on the molecular structure and optical and magnetic properties of fluorinated copper(II) phthalocyanines.

The reduction of copper(ii) octafluoro- {CuII(F8Pc)} and hexadecafluorophthalocyanines {CuII(F16Pc)} by NaCpCo(CO)2 in the presence of cryptand[2.2.2] yields new crystalline {cryptand(Na+)}[CuII(F8Pc)˙3-]-·2C6H4Cl2 (1) and {cryptand(Na+)}2[CuII(F16Pc)4-]2-·C6H14 (2) salts. Together with two previously characterized salts of CuII(FxPc) (x = 8 and 16), this allows the study of the molecular structure and optical and magnetic properties of fluorinated copper phthalocyanines in different reduction states (-1 and -2). The blue shift of the Q-band increases together with the negative charge on the macrocycle, and new weak bands of the anions appear at 820-1013 nm. Alternation of the Nmeso-C bonds manifests itself in reduced macrocycles due to a partial disruption of macrocycle aromaticity. In the case of CuII(F8Pc) this effect is nearly two times stronger for dianions than for monoanions. The alternation of the bonds is less pronounced for perfluorinated CuII(F16Pc)n- (n = 1, 2) anions most probably due to a partial delocalization of the negative charge on fluoro-substituents. The reduction also noticeably elongates the average C-F bonds in CuII(F8Pc). The first reduction centered on the macrocycle leads to the formation of [CuII(F8Pc)˙3-]- in 1 with two S = 1/2 spins positioned on CuII and the radical trianion (F8Pc)˙3- macrocycle. As a result, a broad EPR signal is observed with g = 2.1652 at RT attributable to both paramagnetic species having exchange interactions. The formation of dimerized stacks from [CuII(F8Pc)˙3-]- in 1 results in strong enough magnetic coupling of the (F8Pc)˙3- spins within the dimers (J/kB = -21.8 cm-1), and weaker intramolecular coupling is observed between CuII and (F8Pc)˙3- (J/kB = -10.8 cm-1). Coupling between (F8Pc)˙3- spins from the neighboring dimers is nearly 1.5 times weaker (-14.6 cm-1). Under reduction conditions, a second electron also comes to the macrocycle forming diamagnetic F16Pc4- tetraanions. In this case S = 1/2 spin is preserved on CuII. Magnetic coupling between these centers is weak due to the long distances between them in the [CuII(F8Pc)4-]2- chains of 2. Salt 2 shows an EPR signal with a HF splitting characteristic of CuII with g∥ = 2.1806 (A∥ = 20.11 mT), and g⊥ = 1.9597 at RT.