Peter Pogány

Generative design and machine learning are increasingly prevalent in medicinal chemistry. To pilot the comprehensive use of automated molecular design on a project, the BRADSHAW platform was used to optimize a series of inhibitors of Endoplasmic Reticulum Aminopeptidase 1 (ERAP1), an emerging target in cancer immunotherapy and autoimmune diseases. Through four consecutive iterations applying in silico molecular generation, property prediction and filtering, we conducted a multiparameter optimization of potency, physicochemical properties and pharmacokinetics. Continuous refinement of Machine Learning (ML) models led to improved scoring accuracy and compound quality, culminating in identification of in vitro and in vivo tool molecules. We also discuss our reflections on the pilot and integration of automated design into medicinal chemistry projects, including observations of the human factors resulting from increased use of computational design, and recommendations for future projects.

To understand the benefits and drawbacks of 3D combinatorial and deep learning generators, a novel benchmark was created focusing on the recreation of important protein-ligand interactions and 3D ligand conformations. Using the BindingMOAD data set with a hold-out blind set, the sequential graph neural network generators, Pocket2Mol and PocketFlow, diffusion models, DiffSBDD and MolSnapper, and combinatorial genetic algorithms, AutoGrow4 and LigBuilderV3, were evaluated. It was discovered that deep learning methods fail to generate structurally valid molecules and 3D conformations, whereas combinatorial methods are slow and generate molecules that are prone to failing 2D MOSES filters. The results from this evaluation guide us toward improving deep learning structure-based generators by placing higher importance on structural validity, 3D ligand conformations, and recreation of important known active site interactions. This benchmark should be used to understand the limitations of future combinatorial and deep learning generators. The package is freely available under an Apache 2.0 license at github.com/gskcheminformatics/SBDD-benchmarking.

Predicting reaction yields in synthetic chemistry remains a significant challenge. This study systematically evaluates the impact of tokenization, molecular representation, pretraining data, and adversarial training on a BERT-based model for yield prediction of Buchwald-Hartwig and Suzuki-Miyaura coupling reactions using publicly available HTE data sets. We demonstrate that molecular representation choice (SMILES, DeepSMILES, SELFIES, Morgan fingerprint-based notation, IUPAC names) has minimal impact on model performance, while typically BPE and SentencePiece tokenization outperform other methods. WordPiece is strongly discouraged for SELFIES and fingerprint-based notation. Furthermore, pretraining with relatively small data sets (<100 K reactions) achieves comparable performance to larger data sets containing millions of examples. The use of artificially generated domain-specific pretraining data is proposed. The artificially generated sets prove to be a good surrogate for the reaction schemes extracted from reaction data sets such as Pistachio or Reaxys. The best performance was observed for hybrid pretraining sets combining the real and the domain-specific, artificial data. Finally, we show that a novel adversarial training approach, perturbing input embeddings dynamically, improves model robustness and generalizability for yield and reaction success prediction. These findings provide valuable insights for developing robust and practical machine learning models for yield prediction in synthetic chemistry. GSK's BERT training code base is made available to the community with this work.

We report a ‘direct-to-biology’ (D2B) approach for optimising covalent acrylamide binders of protein targets and apply this to the identification of a selective and cell-active inhibitor of Werner (WRN) helicase. Inhibition of WRN helicase activity exhibits a synthetic lethal relationship with cancers displaying high microsatellite instability (MSI-H) and is being pursued as a therapeutic strategy in the clinic. Using intact-protein liquid chromatography-mass spectrometry (LC-MS) screening, we identified acrylamide fragment binders of the WRN helicase domain and then used covalent D2B chemistry to optimise these initial hits. Our efforts ultimately afforded a potent covalent inhibitor of WRN-mediated DNA unwinding, which displays selective, concentration-dependent cellular engagement of WRN, and demonstrates synthetic lethality in an MSI-H setting. Furthermore, our inhibitor targets a distinct conformation of WRN helicase compared to the current clinical covalent inhibitor, presenting a complementary approach for covalent inhibition of WRN helicase. This work demonstrates how D2B chemistry platforms can be used to explore structure-activity relationships in a modular fashion, while reducing investment of human and material resources.

Predicting reaction yields in synthetic chemistry remains a significant challenge. This study systematically evaluates the impact of tokenization, molecular representation, pre-training data, and adversarial training on a BERT-based model for yield prediction of Buchwald-Hartwig and Suzuki-Miyaura coupling reactions using publicly available HTE datasets. We demonstrate that molecular representation choice (SMILES, DeepSMILES, SELFIES, Morgan fingerprint-based notation, IUPAC names) has minimal impact on model performance, while typically BPE and SentencePiece tokenization outperform other methods. WordPiece is strongly discouraged for SELFIES and fingerprint-based notation. Furthermore, pre-training with relatively small data sets (<100K reactions) achieves comparable performance to larger data sets containing millions of examples. A use of artificially generated domain-specific pre-training data is proposed. The artificially generated sets prove to be a good surrogate to the reaction schemes extracted from reaction data sets such as Pistachio or Reaxys. The best performance was observed for hybrid pre-training sets combining the real and the domain-specific, artificial data. Finally, we show that a novel adversarial training approach, perturbing input embeddings dynamically, improves model robustness and generalisability for yield and reaction success prediction. These findings provide valuable insights for developing robust and practical machine learning models for yield prediction in synthetic chemistry. GSK’s BERT training code base is made available to the community with this work.

A mechanistic study into the copper(I)-catalysed sulfonylative Suzuki–Miyaura reaction, incorporating sulfur dioxide, is described. Utilising spectroscopic and computational techniques, an exploration into the individual components of the competing catalytic cycles is delineated, including identification of the resting state catalyst, transmetalation of arylboronic acid onto copper(I), the sulfur dioxide insertion process, and the oxidative addition of aryl halide to Cu^I. Studies also investigated prominent side-reactions which were uncovered, including a competing copper(II)-catalysed mechanism. This led to an additional proposed and connected Cu^I/Cu^II/Cu^III catalytic cycle to account for by-product formation.

Graph based methods are increasingly important in chemistry and drug discovery, with applications ranging from QSAR to molecular generation. Combining graph neural networks and deep metric learning concepts, we expose a framework for quantifying molecular similarity based on learned embeddings separate from any endpoint. Using a minimal definition of similarity, and data from the ZINC database of public compounds, this work demonstrate the properties of the embedding and its suitability for a range of applications, among them a novel reconstruction loss method for training deep molecular auto-encoders. We also compare the performance of the embedding to standard practices, with a focus on known failure points and edge cases.

Methods for rapid identification of chemical tools are essential for the validation of emerging targets and to provide medicinal chemistry starting points for the development of new medicines. Here, we report a screening platform that combines ‘direct-to-biology’ high-throughput chemistry (D2B-HTC) with photoreactive fragments. The platform enabled the rapid synthesis of >1000 PhotoAffinity Bits (HTC-PhABits) in 384-well plates in 24 h and their subsequent screening as crude reaction products with a protein target without purification. Screening the HTC-PhABit library with carbonic anhydrase I (CAI) afforded 7 hits (0.7% hit rate), which were found to covalently crosslink in the Zn²⁺ binding pocket. A powerful advantage of the D2B-HTC screening platform is the ability to rapidly perform iterative design–make–test cycles, accelerating the development and optimisation of chemical tools and medicinal chemistry starting points with little investment of resource.

Metal–polymer coordination strategy replaced covalent crosslinking, and improved the chemical resistance of membranes without sacrificing their molecular sieving performance.

Machine learning approaches promise to accelerate and improve success rates in medicinal chemistry programmes by more effectively leveraging available data to guide molecular design. A key step of an automated computational design algorithm is molecule generation, where the machine is required to design high-quality, drug-like molecules, within the appropriate chemical space. Many algorithms have been proposed for molecular generation, however a challenge is how to assess the validity of the resulting molecules. Here we report three Turing-inspired tests designed to evaluate the performance of molecular generators. Profound differences were observed between the performance of molecule generators in these tests, highlighting the importance of selection of the appropriate design algorithms for specific circumstances. One molecule generator, based on match-molecular pairs, performed excellently against all tests, and thus provides a valuable component for machine driven medicinal chemistry design workflows.