Metis: a python-based user interface to collect expert feedback for generative chemistry models.

One challenge that current de novo drug design models face is a disparity between the user's expectations and the actual output of the model in practical applications. Tailoring models to better align with chemists' implicit knowledge, expectation and preferences is key to overcoming this obstacle effectively. While interest in preference-based and human-in-the-loop machine learning in chemistry is continuously increasing, no tool currently exists that enables the collection of standardized and chemistry-specific feedback. Metis is a Python-based open-source graphical user interface (GUI), designed to solve this and enable the collection of chemists' detailed feedback on molecular structures. The GUI enables chemists to explore and evaluate molecules, offering a user-friendly interface for annotating preferences and specifying desired or undesired structural features. By providing chemists the opportunity to give detailed feedback, allows researchers to capture more efficiently the chemist's implicit knowledge and preferences. This knowledge is crucial to align the chemist's idea with the de novo design agents. The GUI aims to enhance this collaboration between the human and the "machine" by providing an intuitive platform where chemists can interactively provide feedback on molecular structures, aiding in preference learning and refining de novo design strategies. Metis integrates with the existing de novo framework REINVENT, creating a closed-loop system where human expertise can continuously inform and refine the generative models.Scientific contributionWe introduce a novel Graphical User Interface, that allows chemists/researchers to give detailed feedback on substructures and properties of small molecules. This tool can be used to learn the preferences of chemists in order to align de novo drug design models with the chemist's ideas. The GUI can be customized to fit different needs and projects and enables direct integration into de novo REINVENT runs. We believe that Metis can facilitate the discussion and development of novel ways to integrate human feedback that goes beyond binary decisions of liking or disliking a molecule.

QSARtuna: An Automated QSAR Modeling Platform for Molecular Property Prediction in Drug Design.

Machine-learning (ML) and deep-learning (DL) approaches to predict the molecular properties of small molecules are increasingly deployed within the design-make-test-analyze (DMTA) drug design cycle to predict molecular properties of interest. Despite this uptake, there are only a few automated packages to aid their development and deployment that also support uncertainty estimation, model explainability, and other key aspects of model usage. This represents a key unmet need within the field, and the large number of molecular representations and algorithms (and associated parameters) means it is nontrivial to robustly optimize, evaluate, reproduce, and deploy models. Here, we present QSARtuna, a molecule property prediction modeling pipeline, written in Python and utilizing the Optuna, Scikit-learn, RDKit, and ChemProp packages, which enables the efficient and automated comparison between molecular representations and machine learning models. The platform was developed by considering the increasingly important aspect of model uncertainty quantification and explainability by design. We provide details for our framework and provide illustrative examples to demonstrate the capability of the software when applied to simple molecular property, reaction/reactivity prediction, and DNA encoded library enrichment classification. We hope that the release of QSARtuna will further spur innovation in automatic ML modeling and provide a platform for education of best practices in molecular property modeling. The code for the QSARtuna framework is made freely available via GitHub.

Do Chemformers Dream of Organic Matter? Evaluating a Transformer Model for Multistep Retrosynthesis.

Synthesis planning of new pharmaceutical compounds is a well-known bottleneck in modern drug design. Template-free methods, such as transformers, have recently been proposed as an alternative to template-based methods for single-step retrosynthetic predictions. Here, we trained and evaluated a transformer model, called the Chemformer, for retrosynthesis predictions within drug discovery. The proprietary data set used for training comprised ∼18 M reactions from literature, patents, and electronic lab notebooks. Chemformer was evaluated for the purpose of both single-step and multistep retrosynthesis. We found that the single-step performance of Chemformer was especially good on reaction classes common in drug discovery, with most reaction classes showing a top-10 round-trip accuracy above 0.97. Moreover, Chemformer reached a higher round-trip accuracy compared to that of a template-based model. By analyzing multistep retrosynthesis experiments, we observed that Chemformer found synthetic routes, leading to commercial starting materials for 95% of the target compounds, an increase of more than 20% compared to the template-based model on a proprietary compound data set. In addition to this, we discovered that Chemformer suggested novel disconnections corresponding to reaction templates, which are not included in the template-based model. These findings were further supported by a publicly available ChEMBL compound data set. The conclusions drawn from this work allow for the design of a synthesis planning tool where template-based and template-free models work in harmony to optimize retrosynthetic recommendations.

When Yield Prediction Does Not Yield Prediction: An Overview of the Current Challenges.

Machine Learning (ML) techniques face significant challenges when predicting advanced chemical properties, such as yield, feasibility of chemical synthesis, and optimal reaction conditions. These challenges stem from the high-dimensional nature of the prediction task and the myriad essential variables involved, ranging from reactants and reagents to catalysts, temperature, and purification processes. Successfully developing a reliable predictive model not only holds the potential for optimizing high-throughput experiments but can also elevate existing retrosynthetic predictive approaches and bolster a plethora of applications within the field. In this review, we systematically evaluate the efficacy of current ML methodologies in chemoinformatics, shedding light on their milestones and inherent limitations. Additionally, a detailed examination of a representative case study provides insights into the prevailing issues related to data availability and transferability in the discipline.

In silico rationalisation of selectivity and reactivity in Pd-catalysed C-H activation reactions.

A computational approach has been developed to automatically generate and analyse the structures of the intermediates of palladium-catalysed carbon-hydrogen (C-H) activation reactions as well as to predict the final products. Implemented as a high-performance computing cluster tool, it has been shown to correctly choose the mechanism and rationalise regioselectivity of chosen examples from open literature reports. The developed methodology is capable of predicting reactivity of various substrates by differentiation between two major mechanisms - proton abstraction and electrophilic aromatic substitution. An attempt has been made to predict new C-H activation reactions. This methodology can also be used for the automated reaction planning, as well as a starting point for microkinetic modelling.

Unveiling the role of boroxines in metal-free carbon-carbon homologations using diazo compounds and boronic acids.

By means of computational and experimental mechanistic studies the fundamental role of boroxines in the reaction between diazo compounds and boronic acids was elucidated. Consequently, a selective metal-free carbon-carbon homologation of aryl and vinyl boroxines using TMSCHN2, giving access to TMS-pinacol boronic ester products, was developed.

Visible Light Activation of Boronic Esters Enables Efficient Photoredox C(sp2 )-C(sp3 ) Cross-Couplings in Flow.

We report herein a new method for the photoredox activation of boronic esters. Using these reagents, an efficient and high-throughput continuous flow process was developed to perform a dual iridium- and nickel-catalyzed C(sp2 )-C(sp3 ) coupling by circumventing solubility issues associated with potassium trifluoroborate salts. Formation of an adduct with a pyridine-derived Lewis base was found to be essential for the photoredox activation of the boronic esters. Based on these results we were able to develop a further simplified visible light mediated C(sp2 )-C(sp3 ) coupling method using boronic esters and cyano heteroarenes under flow conditions.

Asymmetric Total Synthesis of (-)-Erogorgiaene and Its C-11 Epimer and Investigation of Their Antimycobacterial Activity.

A short, nine-step, highly enantioselective synthesis of (-)-erogorgiaene and its C-11 epimer is reported. The key stereochemistry controlling steps involve catalytic asymmetric crotylation, anionic oxy-Cope rearrangement and cationic cyclisation. (-)-Erogorgiaene exhibited promising antitubercular activity against multidrug-resistant strains of Mycobacterium tuberculosis.

Continuous-Flow Synthesis and Derivatization of Aziridines through Palladium-Catalyzed C(sp(3) )-H Activation.

A continuous-flow synthesis of aziridines by palladium-catalyzed C(sp(3) )-H activation is described. The new flow reaction could be combined with an aziridine-ring-opening reaction to give highly functionalized aliphatic amines through a consecutive process. A predictive mechanistic model was developed and used to design the C-H activation flow process and illustrates an approach towards first-principles design based on novel catalytic reactions.

Cross-Aldol Reaction of Isatin with Acetone Catalyzed by Leucinol: A Mechanistic Investigation.

Comprehensive mechanistic studies on the enantioselective aldol reaction between isatin (1 a) and acetone, catalyzed by L-leucinol (3 a), unraveled that isatin, apart from being a substrate, also plays an active catalytic role. Conversion of the intermediate oxazolidine 4 into the reactive syn-enamine 6, catalyzed by isatin, was identified as the rate-determining step by both the calculations (ΔG(≠) =26.1 kcal mol(-1) for the analogous L-alaninol, 3 b) and the kinetic isotope effect (kH /kD =2.7 observed for the reaction using [D6 ]acetone). The subsequent reaction of the syn-enamine 6 with isatin produces (S)-2 a (calculated ΔG(≠) =11.6 kcal mol(-1) ). The calculations suggest that the overall stereochemistry is controlled by two key events: 1) the isatin-catalyzed formation of the syn-enamine 6, which is thermodynamically favored over its anti-rotamer 7 by 2.3 kcal mol(-1) ; and 2) the high preference of the syn-enamine 6 to produce (S)-2 a on reaction with isatin (1 a) rather than its enantiomer (ΔΔG(≠) =2.6 kcal mol(-1) ).

Development of a web-based platform for studying lithiation reactions in silico.

A novel integrated web-based system which can compute, visualise and store systematised key parameters of a reaction has been developed from open-source components. As a proof of concept, it has been used to rationalise and predict the regioselectivity of lithiation reactions as well as relative reactivity of substrates. Excellent agreement between the in silico analysis and experimental data was obtained.

Catalytic asymmetric crotylation of aldehydes: application in total synthesis of (-)-elisabethadione.

A new, highly efficient Lewis base catalyst for a practical enantio- and diastereoselective crotylation of unsaturated aldehydes with E- and Z-crotyltrichlorosilanes has been developed. The method was employed as a key step in a novel asymmetric synthesis of bioactive serrulatane diterpene (-)-elisabethadione. Other strategic reactions for setting up the stereogenic centers included anionic oxy-Cope rearrangement and cationic cyclization. The synthetic route relies on simple, high yielding reactions and avoids use of protecting groups or chiral auxiliaries.

Expedient preparation of nazlinine and a small library of indole alkaloids using flow electrochemistry as an enabling technology.

An expedient synthesis of the indole alkaloid nazlinine is reported. Judicious choice of flow electrochemistry as an enabling technology has permitted the rapid generation of a small library of unnatural relatives of this biologically active molecule. Furthermore, by conducting the key electrochemical Shono oxidation in a flow cell, the loading of electrolyte can be significantly reduced to 20 mol % while maintaining a stable, broadly applicable process.

Highly enantio- and diastereoselective generation of two quaternary centers in spirocyclopropanation of oxindole derivatives.

Spirocyclopropanes: Only one out of eight possible stereoisomers was obtained in the asymmetric cascade cyclopropanation of alkylidene oxindoles with ethyl 2-chloroacetoacetate. Improved catalyst design ensured that spirocyclopropyl oxindoles featuring two quaternary centers were synthesized in high yield and high enantio- and diastereoselectivity (see scheme).

Stereoselective palladium-catalyzed functionalization of homoallylic alcohols: a convenient synthesis of di- and trisubstituted isoxazolidines and ß-amino-δ-hydroxy esters.

Enantiopure, Boc-protected alkoxyamines 12 and 13, derived from the readily available homoallylic alcohols 4 via a reaction that involves either inversion or retention of configuration, undergo a diastereoselective Pd-catalyzed ring-closing carbonylative amidation to produce isoxazolidines 16/17 (≤50:1 diastereoisomer ratio (d.r.)) that can be readily converted into the N-Boc-protected esters of ß-amino-δ-hydroxy acids and their γ-substituted homologues 37. The key carbonylative cyclization proceeds through an unusual syn addition of the palladium and the nitrogen nucleophile across the C=C bond (19→21), as revealed by the reaction of 15, which afforded isoxazolidine 18 with high diastereoselectivity.

Direct preparation of thiazoles, imidazoles, imidazopyridines and thiazolidines from alkenes.

A range of heterocycles, namely thiazoles, imidazoles, imidazopyridines, thiazolidines and dimethoxyindoles, have been synthesised directly from alkenes via a two-step ketoidoination/cyclisation protocol. The alkene starting materials are themselves readily accessible using many different and well-established approaches, and allow access to a variety of heterocycles with excellent yields and regioselectivity.

Exerting control over the acyloin reaction.

A synthetic method for conducting the acyloin reaction using electron transfer in solution is reported. By linking two esters via their oxygen atoms, it was possible to perform crossed acyloin reactions between two different ester functionalities and display a high degree of preference for an intramolecular coupling process.

Enantioselective and catalytic method for alpha-crotylation of aldehydes with a kinetic self-refinement of stereochemistry.

Kinetic refinery: A practical, highly stereoselective, two-step catalytic protocol for the alpha-allylation of aldehydes, starting from crotyltrichlorosilanes, has been developed (see scheme). In each reaction step, one of the stereoisomers reacted faster than the other, which resulted in a kinetic stereochemical (E/Z) self-refinement of the system and led to the formation of virtually enantiomerically and geometrically pure linear homoallylic alcohols in high yield.

Vicinal amino alcohols as organocatalysts in asymmetric cross-aldol reaction of ketones: application in the synthesis of convolutamydine A.

Leucinol and valinol have been identified as efficient organocatalysts for the aldol reaction of isatin and its derivatives (as examples of activated, non-enolizable ketones) with acetone. Uncommon mechanistic features were observed and used in the formulation of the transition state of the reaction.