Global reactivity models are impactful in industrial synthesis applications.

Artificial Intelligence is revolutionizing many aspects of the pharmaceutical industry. Deep learning models are now routinely applied to guide drug discovery projects leading to faster and improved findings, but there are still many tasks with enormous unrealized potential. One such task is the reaction yield prediction. Every year more than one fifth of all synthesis attempts result in product yields which are either zero or too low. This equates to chemical and human resources being spent on activities which ultimately do not progress the programs, leading to a triple loss when accounting for the cost of opportunity in time wasted. In this work we pre-train a BERT model on more than 16 million reactions from 4 different data sources, and fine tune it to achieve an uncertainty calibrated global yield prediction model. This model is an improvement upon state of the art not just from the increase in pre-train data but also by introducing a new embedding layer which solves a few limitations of SMILES and enables integration of additional information such as equivalents and molecule role into the reaction encoding, the model is called BERT Enriched Embedding (BEE). The model is benchmarked on an open-source dataset against a state-of-the-art synthesis focused BERT showing a near 20-point improvement in r2 score. The model is fine-tuned and tested on an internal company data benchmark, and a prospective study shows that the application of the model can reduce the total number of negative reactions (yield under 5%) ran in Janssen by at least 34%. Lastly, we corroborate the previous results through experimental validation, by directly deploying the model in an on-going drug discovery project and showing that it can also be used successfully as a reagent recommender due to its fast inference speed and reliable confidence estimation, a critical feature for industry application.

Bidirectional Graphormer for Reactivity Understanding: Neural Network Trained to Reaction Atom-to-Atom Mapping Task.

This work introduces GraphormerMapper, a new algorithm for reaction atom-to-atom mapping (AAM) based on a transformer neural network adopted for the direct processing of molecular graphs as sets of atoms and bonds, as opposed to SMILES/SELFIES sequence-based approaches, in combination with the Bidirectional Encoder Representations from Transformers (BERT) network. The graph transformer serves to extract molecular features that are tied to atoms and bonds. The BERT network is used for chemical transformation learning. In a benchmarking study with IBM RxnMapper, which is the best AAM algorithm according to our previous study, we demonstrate that our AAM algorithm is superior to it on our "Golden" benchmarking data set.

Improving Few- and Zero-Shot Reaction Template Prediction Using Modern Hopfield Networks.

Finding synthesis routes for molecules of interest is essential in the discovery of new drugs and materials. To find such routes, computer-assisted synthesis planning (CASP) methods are employed, which rely on a single-step model of chemical reactivity. In this study, we introduce a template-based single-step retrosynthesis model based on Modern Hopfield Networks, which learn an encoding of both molecules and reaction templates in order to predict the relevance of templates for a given molecule. The template representation allows generalization across different reactions and significantly improves the performance of template relevance prediction, especially for templates with few or zero training examples. With inference speed up to orders of magnitude faster than baseline methods, we improve or match the state-of-the-art performance for top-k exact match accuracy for k ≥ 3 in the retrosynthesis benchmark USPTO-50k. Code to reproduce the results is available at github.com/ml-jku/mhn-react.

Atom-to-atom Mapping: A Benchmarking Study of Popular Mapping Algorithms and Consensus Strategies.

In this paper, we compare the most popular Atom-to-Atom Mapping (AAM) tools: ChemAxon,[1] Indigo,[2] RDTool,[3] NameRXN (NextMove),[4] and RXNMapper[5] which implement different AAM algorithms. An open-source RDTool program was optimized, and its modified version ("new RDTool") was considered together with several consensus mapping strategies. The Condensed Graph of Reaction approach was used to calculate chemical distances and develop the "AAM fixer" algorithm for an automatized correction of erroneous mapping. The benchmarking calculations were performed on a Golden dataset containing 1851 manually mapped and curated reactions. The best performing RXNMapper program together with the AMM Fixer was applied to map the USPTO database. The Golden dataset, mapped USPTO and optimized RDTool are available in the GitHub repository https://github.com/Laboratoire-de-Chemoinformatique.

Synthesis and biological evaluation of 68Ga-bis-DOTA-PA as a potential agent for positron emission tomography imaging of necrosis.

INTRODUCTION: Necrosis is a form of cell death that occurs in a variety of pathological conditions but can also be the result of therapy in cancer treatment. A radiotracer that could image necrotic cell death using PET could therefore be a useful tool to provide relevant information on the disease activity or therapeutic efficacy and assist in diagnosis and therapy management of several disorders. Pamoic acid derivatives have previously been reported to show a selective uptake in tissue undergoing cellular death via necrosis. In this study 4,4'-methylene-bis(2-hydroxy-3-naphthoic hydrazide) (pamoic acid bis-hydrazide) was conjugated to the macrocyclic ligand DOTA and labeled with the generator produced positron emitter (68)Ga. The resulting complex ((68)Ga-bis-DOTA-PA; (68)Ga-3) was evaluated as a potential radiotracer for imaging tissues undergoing cellular death via necrosis. METHODS: Bis-DOTA-PA was synthesized and labeled with (68)Ga. Biodistribution of (68)Ga-3 and analysis of plasma were studied in normal NMRI mice. Binding of the complex to necrotic tissue was first evaluated by in vitro autoradiography. Further evaluation of the uptake in necrotic tissue was performed in two different models of necrosis using microPET imaging in correlation with ex vivo autoradiography, biodistribution studies and histochemical staining. A biodistribution study in a mouse model of hepatic apoptosis was performed to study the selectivity of the uptake of (68)Ga-bis-DOTA-PA in necrotic tissue. RESULTS: (68)Ga-3 was obtained with a decay-corrected radiochemical yield of 51.8% ± 5.4% and a specific activity of about 12 GBq/µmol. In normal mice, the complex was slowly cleared from blood, mainly through the renal pathway, and showed high in vivo stability. (68)Ga-bis-DOTA-PA displayed high and selective uptake in necrotic tissue and allowed imaging of necrotic tissue using microPET. CONCLUSION: (68)Ga-3 was synthesized and characterized. In vitro, in vivo and ex vivo studies showed that the complex displays high and selective uptake in tissue undergoing cellular death via necrosis.

Synthesis and biological evaluation of 68Ga labeled bis-DOTA-3,3'-(benzylidene)-bis-(1H-indole-2-carbohydrazide) as a PET tracer for in vivo visualization of necrosis.

The aim of the present study was to develop a (68)Ga labeled bis-DOTA derivative of benzylidene-bis-indole and compare the in vivo stability and biodistribution with that of the previously reported bis-DTPA derivate for in vivo imaging of necrosis using PET. Uptake of the tracer was evaluated in a mouse model of Fas-mediated hepatic apoptosis in correlation with histochemical stainings. The novel (68)Ga labeled tracer showed an improved in vivo stability and could therefore be used for selective non-invasive imaging of necrotic cell death using PET.

Sedoheptulose accumulation under CO2 enrichment in leaves of Kalanchoë pinnata: a novel mechanism to enhance C and P homeostasis?

In contrast to the well-documented roles of its mono- and bisphosphate esters, the occurrence of free sedoheptulose in plant tissues remains a matter of conjecture. The present work sought to determine the origin of sedoheptulose formation in planta, as well as its physiological importance. Elevated CO2 and sucrose induction experiments were used to study sedoheptulose metabolism in the Crassulacean acid metabolism (CAM) plants Kalanchoë pinnata and Sedum spectabile. Experimental evidence suggested that sedoheptulose is produced from the oxidative pentose phosphate pathway intermediate sedoheptulose-7-phosphate, by a sedoheptulose-7-phosphate phosphatase. Carbon flux through this pathway was stimulated by increased triose-phosphate levels (elevated CO2, compromised sink availability, and sucrose incubation of source leaves) and attenuated by ADP and inorganic phosphate (Pi). The accumulation of free sedoheptulose is proposed to act as a mechanism contributing to both C and P homeostasis by serving as an alternative carbon store under elevated CO2 or a compromised sink capacity to avoid sucrose accumulation, depletion of inorganic phosphate, and suppression of photosynthesis. It remains to be established whether this acclimation-avoiding mechanism is confined to CAM plants, which might be especially vulnerable to Pi imbalances, or whether some C3 and C4 plants also dispose of the genetic capacity to induce and accelerate sedoheptulose synthesis upon CO2 elevation.

Design of bioactive peptides from naturally occurring µ-conotoxin structures.

To date, cone snail toxins ("conotoxins") are of great interest in the pursuit of novel subtype-selective modulators of voltage-gated sodium channels (Na(v)s). Na(v)s participate in a wide range of electrophysiological processes. Consequently, their malfunctioning has been associated with numerous diseases. The development of subtype-selective modulators of Na(v)s remains highly important in the treatment of such disorders. In current research, a series of novel, synthetic, and bioactive compounds were designed based on two naturally occurring µ-conotoxins that target Na(v)s. The initial designed peptide contains solely 13 amino acids and was therefore named "Mini peptide." It was derived from the µ-conotoxins KIIIA and BuIIIC. Based on this Mini peptide, 10 analogues were subsequently developed, comprising 12-16 amino acids with two disulfide bridges. Following appropriate folding and mass verification, blocking effects on Na(v)s were investigated. The most promising compound established an IC(50) of 34.1 ± 0.01 nM (R2-Midi on Na(v)1.2). An NMR structure of one of our most promising compounds was determined. Surprisingly, this structure does not reveal an α-helix. We prove that it is possible to design small peptides based on known pharmacophores of µ-conotoxins without losing their potency and selectivity. These data can provide crucial material for further development of conotoxin-based therapeutics.

Reactivity of amino acid nucleoside phosphoramidates: a mechanistic quantum chemical study.

Recent experimental evidence (Maiti et al. Chem.-Eur. J., submitted) indicates that hydrolysis of nucleoside phosphoramidates is subjected to anchimeric influence by carboxyl moieties in the leaving group but also by the base in the nucleotide. A quantum chemical analysis of these findings is presented. First the intrinsic hydrolysis mechanism is investigated for simplified model compounds, and then both amino acid and nucleoside substituents are included. It is found that hydrolysis is assisted by the α-carboxyl group via formation of a five-membered intermediate and that the barrier for the reaction of this intermediate toward the product state can be influenced by the nucleobase. The adenine base protonated on N3 interacts with the transition state and considerably lowers the barrier for hydrolysis. The influence of several base modifications is explained by calculating the pK(a) for protonation on N3.

Pc16a, the first characterized peptide from Conus pictus venom, shows a novel disulfide connectivity.

A novel conotoxin, pc16a, was isolated from the venom of Conus pictus. This is the first peptide characterized from this South-African cone snail and it has only 11 amino acid residues, SCSCKRNFLCC*, with the rare cysteine framework XVI and a monoisotopic mass of 1257.6Da. Two peptides were synthesized with two possible conformations: globular (pc16a_1) and ribbon (pc16a_2). pc16a_1 co-eluted with the native peptide, which indicates a disulfide connectivity I-III, II-IV. The structure of pc16a_1 was determined by NMR. Both synthetic peptides were used to elucidate the biological activity. Bioassays were performed on crickets, ghost shrimps, larvae of the mealworm beetle and mice, but no effect was seen. Using two-electrode voltage clamp, a range of voltage-gated ion channels (Na(v) and K(v)) and nicotinic acetylcholine receptors were screened, but again no activity was found. Hence, the specific target of pc16a still remains to be discovered.

Influence of the nucleobase and anchimeric assistance of the carboxyl acid groups in the hydrolysis of amino acid nucleoside phosphoramidates.

Nucleoside phosphoramidates (NPs) are a class of nucleotide analogues that has been developed as potential antiviral/antitumor prodrugs. Recently, we have shown that some amino acid nucleoside phosphoramidates (aaNPs) can act as substrates for viral polymerases like HIV-1 RT. Herein, we report the synthesis and hydrolysis of a series of new aaNPs, containing either natural or modified nucleobases to define the basis for their differential reactivity. Aqueous stability, kinetics, and hydrolysis pathways were studied by NMR spectroscopy at different solution pD values (5-7) and temperatures. It was observed that the kinetics and mechanism (P-N and/or P-O bond cleavage) of the hydrolysis reaction largely depend on the nature of the nucleobase and amino acid moieties. Aspartyl NPs were found to be more reactive than Gly or ß-Ala NPs. For aspartyl NPs, the order of reactivity of the nucleobase was 1-deazaadenine>7-deazaadenine>adenine>thymine≥3-deazaadenine. Notably, neutral aqueous solutions of Asp-1-deaza-dAMP degraded spontaneously even at 4 °C through exclusive P-O bond hydrolysis (a 50-fold reactivity difference for Asp-1-deaza-dAMP vs. Asp-3-deaza-dAMP at pD 5 and 70 °C). Conformational studies by NMR spectroscopy and molecular modeling suggest the involvement of the protonated N3 atom in adenine and 1- and 7-deazaadenine in the intramolecular catalysis of the hydrolysis reaction through the rare syn conformation.

Structure determination of the top-loop of the conserved 3'-terminal secondary structure in the genome of flaviviruses.

To what extent small differences in RNA sequences (mutations) can have a profound impact on biology remains an intriguing question. This effect can be studied by using untranslated RNA regions as a model. We have studied the influence of mutations on the structure of an RNA hairpin that occurs in the 3'-untranslated region (UTR) of Flaviviridae, and is known to have a large impact on the vector dependency of flaviviruses. Three related RNA sequences were studied by NMR spectroscopy. The selected sequences represent each one of the three clusters in the flavivirus genes (mosquito-borne, tick-borne, and no-known-vector viruses). A new strategy was used to obtain chemical shift signatures of carbonyl atoms in unlabeled uridine nucleobases to characterize their involvement in hydrogen bonding. Clear differences occur in the structures and stacking pattern of the three RNA hairpins. The observed differences cannot be predicted based on sequence analysis. A different biology can be correlated with a different RNA tertiary structure. The underlying biological mechanism, however, remains to be studied.

Metabolism of galactosyl-oligosaccharides in Stellaria media--discovery of stellariose synthase, a novel type of galactosyltransferase.

The raffinose family oligosaccharides (RFOs), including raffinose (Gal-alpha(1-->6)-Glc-alpha(1-->2)beta-Fru), stachyose (Gal-alpha(1-->6)-Gal-alpha(1-->6)-Glc-alpha(1-->2)beta-Fru) and higher degree of polymerization RFOs are the most widespread galactosyl-oligosaccharides (GOS) in the plant kingdom. Stellaria media is a typical representative of the Caryophyllaceae, a plant family lacking stachyose and the typical galactosyl extensions of stachyose. During cold treatment raffinose, lychnose (Gal-alpha(1-->6)-Glc-alpha(1-->2)beta-Fru-alpha(1-->1)-Gal) and stellariose (Gal-alpha(1-->6)-[Gal-alpha(1-->4)]-Glc-alpha(1-->2)beta-Fru-alpha(1-->1)-Gal) were found to accumulate in S. media stems. Next to these prominent oligosaccharides, two extra GOS were discovered. Biochemical analyses (enzymatic incubations and mild acid hydrolysis) and mass spectrometry identified the first, most abundant oligosaccharide as Glc-alpha(1-->2)beta-Fru-alpha(1-->1)-Gal, a breakdown product of lychnose. The structure of this trisaccharide was confirmed by full NMR characterization. The second, less abundant compound (termed mediose) was identified as Gal-alpha(1-->6)-[Gal-alpha(1-->4)]Glc-alpha(1-->2)beta-Fru after biochemical analyses. By partial enzyme purification the presence of discrete lychnose synthase (raffinose:raffinose 1(Fru) galactosyltransferase) and stellariose synthase (raffinose:lychnose 4(Glc) galactosyltransferase) activities were shown. A model is presented explaining the structural diversity of GOS in S. media. In the absence of stachyose, raffinose is further elongated by lychnose synthase and stellariose synthase to produce lychnose, mediose and stellariose. Most likely, these compounds are also subject to partial trimming by endogenous alpha-galactosidases.

Phosphodiester substrates for incorporation of nucleotides in DNA using HIV-1 reverse transcriptase.

In previous research we demonstrated that some amino acid derivatives of deoxyadenosine 5'-O-monophosphate act as substrates for incorporation into DNA by HIV-1 reverse transcriptase while retaining the canonical base-pair selectivity for all natural bases. Thus, some amino acids mimic the pyrophosphate group in the polymerization process with this enzyme. Herein we extended this study to the evaluation of a range of potential new leaving groups with aromatic and aliphatic structures carrying one or two carboxylic acid functions. Out of this series, the isophthalic acid derivative of deoxyadenosine 5'-O-monophosphate gave single-nucleotide incorporation results similar to those obtained with the L-aspartic acid phosphoramidate of deoxyadenosine monophosphate. The glycolic acid analogue is a better substrate than the glycine congener, supporting the good leaving group properties of a phosphodiester linkage. These investigations provide new insight into the structural requirements for leaving groups that can mimic the pyrophosphate moiety of nucleoside triphosphates.

Isolation and characterization of a pentasaccharide from Stellaria media.

While classic raffinose family oligosaccharides (RFOs) such as raffinose and stachyose are common in plants, stachyose is absent in the Caryophyllaceae. Instead the tetrasaccharide lychnose α-d-Gal-(1→6)α-d-Glc-(1→2)ß-d-Fru-(1→1)α-d-Gal can accumulate. Stellaria media, a representative member of this family, was used to isolate α-d-Gal-(1→6)-[α-d-Gal-(1→4)]α-d-Glc-(1→2)ß-d-Fru-(1→1)α-d-Gal, a novel pentasaccharide with a lychnose backbone. Complete NMR characterization using COSY, HSQC, HSQC-TOCSY, HMBC, and NOESY experiments was performed to unequivocally resolve its structure. This is the first report of a natural compound containing a Gal α(1→4)Glc linkage. The trivial name stellariose is proposed for this new pentasaccharide.

Synthesis and biological evaluation of an 123I-labeled bicyclic nucleoside analogue (BCNA) as potential SPECT tracer for VZV-tk reporter gene imaging.

An iodine-123 labeled bicyclic nucleoside analogue ([(123)I]-4) has been synthesized and evaluated as a potential single photon emission tomography (SPECT) reporter probe for the non-invasive imaging of expression of the varicella zoster virus thymidine kinase (VZV-tk) reporter gene. In vitro enzymatic assays revealed that the non-radioactive mono-iodo derivative 4 has good affinity for VZV-TK (IC(50): 4.2 microM). Biodistribution of [(123)I]-4 was examined in normal mice. Evaluation of [(123)I]-4 in HEK-293T cells showed 1.74-fold higher accumulation in VZV-TK-expressing cells compared to control cells.