Fast Customization of Chemical Language Models to Out-of-Distribution Data Sets.

The world is on the verge of a new industrial revolution, and language models are poised to play a pivotal role in this transformative era. Their ability to offer intelligent insights and forecasts has made them a valuable asset for businesses seeking a competitive advantage. The chemical industry, in particular, can benefit significantly from harnessing their power. Since 2016 already, language models have been applied to tasks such as predicting reaction outcomes or retrosynthetic routes. While such models have demonstrated impressive abilities, the lack of publicly available data sets with universal coverage is often the limiting factor for achieving even higher accuracies. This makes it imperative for organizations to incorporate proprietary data sets into their model training processes to improve their performance. So far, however, these data sets frequently remain untapped as there are no established criteria for model customization. In this work, we report a successful methodology for retraining language models on reaction outcome prediction and single-step retrosynthesis tasks, using proprietary, nonpublic data sets. We report a considerable boost in accuracy by combining patent and proprietary data in a multidomain learning formulation. This exercise, inspired by a real-world use case, enables us to formulate guidelines that can be adopted in different corporate settings to customize chemical language models easily.

A dispensable paralog of succinate dehydrogenase subunit C mediates standing resistance towards a subclass of SDHI fungicides in Zymoseptoria tritici.

Succinate dehydrogenase inhibitor (SDHI) fungicides are widely used for the control of a broad range of fungal diseases. This has been the most rapidly expanding fungicide group in terms of new molecules discovered and introduced for agricultural use over the past fifteen years. A particular pattern of differential sensitivity (resistance) to the stretched heterocycle amide SDHIs (SHA-SDHIs), a subclass of chemically-related SDHIs, was observed in naïve Zymoseptoria tritici populations not previously exposed to these chemicals. Subclass-specific resistance was confirmed at the enzyme level but did not correlate with the genotypes of the succinate dehydrogenase (SDH) encoding genes. Mapping and characterization of the molecular mechanisms responsible for standing SHA-SDHI resistance in natural field isolates identified a gene paralog of SDHC, termed ZtSDHC3, which encodes for an alternative C subunit of succinate dehydrogenase, named alt-SDHC. Using reverse genetics, we showed that alt-SDHC associates with the three other SDH subunits, leading to a fully functional enzyme and that a unique Qp-site residue within the alt-SDHC protein confers SHA-SDHI resistance. Enzymatic assays, computational modelling and docking simulations for the two SQR enzymes (altC-SQR, WT_SQR) enabled us to describe enzyme-inhibitor interactions at an atomistic level and to propose rational explanations for differential potency and resistance across SHA-SDHIs. European Z. tritici populations displayed a presence (20-30%) / absence polymorphism of ZtSDHC3, as well as differences in ZtSDHC3 expression levels and splicing efficiency. These polymorphisms have a strong impact on SHA-SDHI resistance phenotypes. Characterization of the ZtSDHC3 promoter in European Z. tritici populations suggests that transposon insertions are associated with the strongest resistance phenotypes. These results establish that a dispensable paralogous gene determines SHA-SDHIs fungicide resistance in natural populations of Z. tritici. This study paves the way to an increased awareness of the role of fungicidal target paralogs in resistance to fungicides and demonstrates the paramount importance of population genomics in fungicide discovery.

Screening a protein kinase inhibitor library against Plasmodium falciparum.

BACKGROUND: Protein kinases have been shown to be key drug targets, especially in the area of oncology. It is of interest to explore the possibilities of protein kinases as a potential target class in Plasmodium spp., the causative agents of malaria. However, protein kinase biology in malaria is still being investigated. Therefore, rather than assaying against individual protein kinases, a library of 4731 compounds with protein kinase inhibitor-like scaffolds was screened against the causative parasite, Plasmodium falciparum. This approach is more holistic and considers the whole kinome, making it possible to identify compounds that inhibit more than one P. falciparum protein kinase, or indeed other malaria targets. RESULTS: As a result of this screen, 9 active compound series were identified; further validation was carried out on 4 of these series, with 3 being progressed into hits to lead chemistry. The detailed evaluation of one of these series is described. DISCUSSION: This screening approach proved to be an effective way to identify series for further optimisation against malaria. Compound optimisation was carried out in the absence of knowledge of the molecular target. Some of the series had to be halted for various reasons. Mode of action studies to find the molecular target may be useful when problems prevent further chemical optimisation. CONCLUSIONS: Progressible series were identified through phenotypic screening of a relatively small focused kinase scaffold chemical library.

Development of Small-Molecule Trypanosoma brucei N-Myristoyltransferase Inhibitors: Discovery and Optimisation of a Novel Binding Mode.

The enzyme N-myristoyltransferase (NMT) from Trypanosoma brucei has been validated both chemically and biologically as a potential drug target for human African trypanosomiasis. We previously reported the development of some very potent compounds based around a pyrazole sulfonamide series, derived from a high-throughput screen. Herein we describe work around thiazolidinone and benzomorpholine scaffolds that were also identified in the screen. An X-ray crystal structure of the thiazolidinone hit in Leishmania major NMT showed the compound bound in the previously reported active site, utilising a novel binding mode. This provides potential for further optimisation. The benzomorpholinone was also found to bind in a similar region. Using an X-ray crystallography/structure-based design approach, the benzomorpholinone series was further optimised, increasing activity against T. brucei NMT by >1000-fold. A series of trypanocidal compounds were identified with suitable in vitro DMPK properties, including CNS exposure for further development. Further work is required to increase selectivity over the human NMT isoform and activity against T. brucei.

A novel multiple-stage antimalarial agent that inhibits protein synthesis.

There is an urgent need for new drugs to treat malaria, with broad therapeutic potential and novel modes of action, to widen the scope of treatment and to overcome emerging drug resistance. Here we describe the discovery of DDD107498, a compound with a potent and novel spectrum of antimalarial activity against multiple life-cycle stages of the Plasmodium parasite, with good pharmacokinetic properties and an acceptable safety profile. DDD107498 demonstrates potential to address a variety of clinical needs, including single-dose treatment, transmission blocking and chemoprotection. DDD107498 was developed from a screening programme against blood-stage malaria parasites; its molecular target has been identified as translation elongation factor 2 (eEF2), which is responsible for the GTP-dependent translocation of the ribosome along messenger RNA, and is essential for protein synthesis. This discovery of eEF2 as a viable antimalarial drug target opens up new possibilities for drug discovery.

The design and synthesis of potent and selective inhibitors of Trypanosoma brucei glycogen synthase kinase 3 for the treatment of human african trypanosomiasis.

Glycogen synthase kinase 3 (GSK3) is a genetically validated drug target for human African trypanosomiasis (HAT), also called African sleeping sickness. We report the synthesis and biological evaluation of aminopyrazole derivatives as Trypanosoma brucei GSK3 short inhibitors. Low nanomolar inhibitors, which had high selectivity over the off-target human CDK2 and good selectivity over human GSK3ß enzyme, have been prepared. These potent kinase inhibitors demonstrated low micromolar levels of inhibition of the Trypanosoma brucei brucei parasite grown in culture.

Current challenges and trends in the discovery of agrochemicals.

Crop protection chemistry has come a long way from its "alchemic" beginnings in the late 19th century to a high-tech science that supports the sustainable production of food, feed, and fiber for a rapidly growing population. Cutting-edge developments in the design and synthesis of agrochemicals help to tackle today's challenges of weed and pest resistance, higher regulatory safety margins, and higher cost of goods with the invention of selective, environmentally benign, low use rate, and cost-effective active ingredients.

From on-target to off-target activity: identification and optimisation of Trypanosoma brucei GSK3 inhibitors and their characterisation as anti-Trypanosoma brucei drug discovery lead molecules.

Human African trypanosomiasis (HAT) is a life-threatening disease with approximately 30 000-40 000 new cases each year. Trypanosoma brucei protein kinase GSK3 short (TbGSK3) is required for parasite growth and survival. Herein we report a screen of a focused kinase library against T. brucei GSK3. From this we identified a series of several highly ligand-efficient TbGSK3 inhibitors. Following the hit validation process, we optimised a series of diaminothiazoles, identifying low-nanomolar inhibitors of TbGSK3 that are potent in vitro inhibitors of T. brucei proliferation. We show that the TbGSK3 pharmacophore overlaps with that of one or more additional molecular targets.

Mutagenesis and functional studies with succinate dehydrogenase inhibitors in the wheat pathogen Mycosphaerella graminicola.

A range of novel carboxamide fungicides, inhibitors of the succinate dehydrogenase enzyme (SDH, EC 1.3.5.1) is currently being introduced to the crop protection market. The aim of this study was to explore the impact of structurally distinct carboxamides on target site resistance development and to assess possible impact on fitness. We used a UV mutagenesis approach in Mycosphaerella graminicola, a key pathogen of wheat to compare the nature, frequencies and impact of target mutations towards five subclasses of carboxamides. From this screen we identified 27 amino acid substitutions occurring at 18 different positions on the 3 subunits constituting the ubiquinone binding (Qp) site of the enzyme. The nature of substitutions and cross resistance profiles indicated significant differences in the binding interaction to the enzyme across the different inhibitors. Pharmacophore elucidation followed by docking studies in a tridimensional SDH model allowed us to propose rational hypotheses explaining some of the differential behaviors for the first time. Interestingly all the characterized substitutions had a negative impact on enzyme efficiency, however very low levels of enzyme activity appeared to be sufficient for cell survival. In order to explore the impact of mutations on pathogen fitness in vivo and in planta, homologous recombinants were generated for a selection of mutation types. In vivo, in contrast to previous studies performed in yeast and other organisms, SDH mutations did not result in a major increase of reactive oxygen species levels and did not display any significant fitness penalty. However, a number of Qp site mutations affecting enzyme efficiency were shown to have a biological impact in planta.Using the combined approaches described here, we have significantly improved our understanding of possible resistance mechanisms to carboxamides and performed preliminary fitness penalty assessment in an economically important plant pathogen years ahead of possible resistance development in the field.

Identification of inhibitors of the Leishmania cdc2-related protein kinase CRK3.

New drugs are urgently needed for the treatment of tropical parasitic diseases such as leishmaniasis and human African trypanosomiasis (HAT). This work involved a high-throughput screen of a focussed kinase set of ~3400 compounds to identify potent and parasite-selective inhibitors of an enzymatic Leishmania CRK3-cyclin 6 complex. The aim of this study is to provide chemical validation that Leishmania CRK3-CYC6 is a drug target. Eight hit series were identified, of which four were followed up. The optimisation of these series using classical SAR studies afforded low-nanomolar CRK3 inhibitors with significant selectivity over the closely related human cyclin dependent kinase CDK2.

Nucleophilic catalysis of acylhydrazone equilibration for protein-directed dynamic covalent chemistry.

Dynamic covalent chemistry uses reversible chemical reactions to set up an equilibrating network of molecules at thermodynamic equilibrium, which can adjust its composition in response to any agent capable of altering the free energy of the system. When the target is a biological macromolecule, such as a protein, the process corresponds to the protein directing the synthesis of its own best ligand. Here, we demonstrate that reversible acylhydrazone formation is an effective chemistry for biological dynamic combinatorial library formation. In the presence of aniline as a nucleophilic catalyst, dynamic combinatorial libraries equilibrate rapidly at pH 6.2, are fully reversible, and may be switched on or off by means of a change in pH. We have interfaced these hydrazone dynamic combinatorial libraries with two isozymes from the glutathione S-transferase class of enzyme, and observed divergent amplification effects, where each protein selects the best-fitting hydrazone for the hydrophobic region of its active site.

Pyrrolidine derivatives as plasmepsin inhibitors: binding mode analysis assisted by molecular dynamics simulations of a highly flexible protein.

Plasmepsins II (EC number: 3.4.23.39) and IV (EC number: 3.4.23.B14) are aspartic proteases present in the food vacuole of the malaria parasite Plasmodium falciparum and are involved in host hemoglobin degradation. A series of pyrrolidine derivatives, originally synthesized as HIV-1 protease inhibitors, were tested for activity against plasmepsin (Plm). Inhibitors in the nanomolar range were discovered for the Plm II and IV isoforms. Detailed studies were carried out to identify putative binding modes that help to explain the underlying structure-activity relationships. Reasonable binding modes were generated for pyrrolidine-3,4-diester derivatives and a substituted 3,4-diaminopyrrolidine inhibitor by using a crystal structure of inhibitor-bound Plm II (PDB ID: 1LEE). Modeling studies indicated that the flap of available Plm crystal structures is not sufficiently opened to accommodate the 3,4-bis(aminomethylene)pyrrolidines. Molecular dynamics simulations were performed to analyze the flexibility of the protein in greater detail, leading to a binding mode hypothesis for the 3,4-bis(aminomethylene)pyrrolidines and providing further insight and general implications for the design of Plm II inhibitors.

Achiral oligoamines as versatile tool for the development of aspartic protease inhibitors.

Due to the important role that aspartic proteases play in many patho-physiological processes, they have intensively been targeted by modern drug development. However, up to now, only for two family members, renin and HIV protease, approved drugs are available. Inhibitor development, mostly guided by mimicking the natural peptide substrates, resulted in very potent inhibitors for several targets, but the pharmacokinetic properties of these compounds were often not optimal. Herein we report a novel approach for lead structure discovery of non-peptidic aspartic protease inhibitors using easily accessible achiral linear oligoamines as starting point. An initial library comprising 11 inhibitors was developed and screened against six selected aspartic proteases. Several hits could be identified, among them selective as well as rather promiscuous inhibitors. The design concept was confirmed by determination of the crystal structure of two derivatives in complex with the HIV-1 protease, and represents a promising basis for the further inhibitor development.

Computer-aided design and synthesis of nonpeptidic plasmepsin II and IV inhibitors.

Plasmepsins (Plm) II (EC number: 3.4.23.39) and IV (EC number: 3.4.23.B14) are aspartic proteases present in the food vacuole of the malaria parasite Plasmodium falciparum and are involved in host hemoglobin degradation. Based on our established efficient synthetic sequence, a series of inhibitors for Plm II and IV has been synthesized bearing a 2,3,4,7-tetrahydro-1H-azepine scaffold as the core structural element. During the computational design cycle, thorough investigations were carried out in order to find a reasonable theoretical binding mode for Plm II and IV. The conformation of Plm II in the crystal structure (PDB code: 1LF2) provides a good starting geometry for our virtual screening approach. In contrast, the only available co-crystal structure for Plm IV of P. falciparum (PDB code: 1LS5) appears inappropriate for inhibitor design. Therefore, a homology model was constructed based on the Plm II 1LF2 structure. A combinatorial docking run using FlexX(c) suggested compounds which, after synthesis, turned out to exhibit affinities in the sub-micromolar range. The observed structure-activity relationships of the synthesized compounds confirm the assumed binding mode for Plm II and IV. The best-binding inhibitors designed for Plm II and IV are devoid of any inhibitory potency against human cathepsin D (EC number: 3.4.23.5).

The potential of P1 site alterations in peptidomimetic protease inhibitors as suggested by virtual screening and explored by the use of C-C-coupling reagents.

A synthetic concept is presented that allows the construction of peptide isostere libraries through polymer-supported C-acylation reactions. A phosphorane linker reagent is used as a carbanion equivalent; by employing MSNT as a coupling reagent, the C-acylation can be conducted without racemization. Diastereoselective reduction was effected with L-selectride. The reagent linker allows the preparation of a norstatine library with full variation of the isosteric positions including the P1 side chain that addresses the protease S1 pocket. Therefore, the concept was employed to investigate the P1 site specificity of peptide isostere inhibitors systematically. The S1 pocket of several aspartic proteases including plasmepsin II and cathepsin D was modeled and docked with approximately 500 amino acid side chains. Inspired by this virtual screen, a P1 site mutation library was designed, synthesized, and screened against three aspartic proteases (plasmepsin II, HIV protease, and cathepsin D). The potency of norstatine inhibitors was found to depend strongly on the P1 substituent. Large, hydrophobic residues such as biphenyl, 4-bromophenyl, and 4-nitrophenyl enhanced the inhibitory activity (IC50) by up to 70-fold against plasmepsin II. In addition, P1 variation introduced significant selectivity, as up to 9-fold greater activity was found against plasmepsin II relative to human cathepsin D. The active P1 site residues did not fit into the crystal structure; however, molecular dynamics simulation suggested a possible alternative binding mode.