Abstracts & Pre-Reading Material
Some speakers have provided the following abstracts and references which might be of interest ahead of their talks.
Val Gillet, University of Sheffield
Navigating synthetically accessible chemical space
- Ghiandoni GM, Bodkin MJ, Chen B, Hristozov D, Wallace JEA, Webster J & Gillet VJ (2021) RENATE : a pseudo-retrosynthetic tool for synthetically accessible de novo design. Molecular Informatics. https://doi.org/10.1002/minf.202100207
- Ghiandoni G, Bodkin MJ, Chen B, Hristozov D, Wallace JEA, Webster J & Gillet V (2020) Enhancing reaction-based de novo design using a multi-label reaction class recommender. Journal of Computer-Aided Molecular Design. 34, pages783–803. https://doi.org/10.1007/s10822-020-00300-6
- Ghiandoni GM, Bodkin MJ, Chen B, Hristozov D, Wallace JEA, Webster J & Gillet V (2019) Development and Application of a Data-Driven Reaction Classification Model: Comparison of an Electronic Lab Notebook and Medicinal Chemistry Literature. Journal of Chemical Information and Modeling, 59(10), 4167-4187. https:// doi.org/10.1021/acs.jcim.9b00537
- Patel H, Gillet VJ, Chen B & Bodkin MJ (2009) Knowledge-based approach to de Novo design using reaction vectors. Journal of Chemical Information and Modeling, 49(5), 1163-1184. https:// doi.org/10.1021/ci800413m
Franca Klingler, MSD
Chemical Space Docking: Novel ROCK1 Kinase Inhibitors found by Large-Scale Structure-Based Virtual Screening
We present Chemical Space Docking, a novel virtual screening method and its application on ROCK1 kinase. The approach combines two distinct advances: (1) it avoids full library enumeration, (2) products are evaluated by molecular docking, which leverages protein structural information. To our knowledge, this is the only structure-based virtual screening technique that effectively facilitates mining billions of molecules. We applied Chemical Space Docking to identify inhibitors of ROCK1 kinase from almost one billion commercially available synthesis-on-demand compounds. From 69 synthesized molecules, 39% had Ki values below 10 µM. Two leads were crystallized with the ROCK1 protein, and the structures showed excellent agreement with the docking poses. Our approach scales roughly with the number of building blocks that span a chemical space and is therefore multiple orders of magnitude faster than traditional docking of fully enumerated libraries.
Iva Lukac, Charnwood Molecular
QM-SAR: Quantum Mechanics Structure-Activity Relationship
Protein-ligand binding affinities are driven by the balance between multiple factors, many of which can only partially be accounted for by force field-based methods. On the contrary, the first principle nature of quantum mechanics (QM) calculations enables systematic improvements to the accuracy by which biomolecular recognition is described. Despite the greater accuracy, QM methods have not been routinely used in drug discovery due to the high computational costs required to deal with large biological systems. The fragment molecular orbital (FMO) method is an approach in which QM calculations are performed on fragments, thus enabling a high level of accuracy with very high efficiency. Pairwise interaction energy (PIE) between any fragment pairs can then be decomposed into four energy terms: electrostatics, exchange-repulsion, charge transfer, and dispersion, offering an unprecedented insight into the nature of protein-ligand binding. This talk will show examples in which the FMO method was applied to different problems in the early stages of drug discovery, from assessing water energetics, to helping the design of macrocycles, fragment growing, and lead optimisation.
Heifetz A, James T, Southey M, Bodkin MJ, Bromidge S. Guiding Medicinal Chemistry with Fragment Molecular Orbital (FMO) Method. Methods Mol Biol. 2020;2114:37-48, doi: https://doi.org/10.1007/978-1-0716-0282-9_3. PMID: 32016885.
Heifetz A, Sladek V, Townsend-Nicholson A, Fedorov DG. Characterizing Protein-Protein Interactions with the Fragment Molecular Orbital Method. Methods Mol Biol. 2020;2114:187-205, doi: https://doi.org/10.1007/978-1-0716-0282-9_13. PMID: 32016895.
Lukac, I., Wyatt, P.G., Gilbert, I.H. et al. Ligand binding: evaluating the contribution of the water molecules network using the Fragment Molecular Orbital method. J Comput Aided Mol Des 35, 1025–1036 (2021). https://doi.org/10.1007/s10822-021-00416-3
Stefania Monteleone, Evotec
Identification of PPI Hotspots and Modulators Using the FMO-PPI Method
Protein functions and signalling are mediated by protein-protein interactions (PPIs) and the identification of the key interacting residues at the interface (hotspots) is the first step for the design of PPI modulators. A fast and accurate way to obtain a list of interactions between key residues, including their chemical nature (electrostatic or hydrophobic) and strength (in kcal/mol) is the Fragment Molecular Orbital (FMO) method.We combined FMO and PPI exploration in a new workflow (FMO-PPI1) to identify not only the PPI hotspots, but also the intramolecular interactions and significant water bridges that stabilize the interface. We benchmarked FMO-PPI with a dataset of protein-protein complexes that represent different protein subfamilies and compared its outcome to published site directed mutagenesis data. We also showed that FMO-PPI can be used to support structure-based drug design of PPI modulators. Here we will present examples of its application to the hit-to-lead and lead optimisation phases of PPI inhibitors and molecular glues.
Daniel Mason and Daniel O’Donovan, HealX
Born of Mind and Machine: Augmenting the drug repositioning process with omics- and graph based machine learning
There are 7,000 known rare diseases that affect 400 million people across the globe, and 95% of them have no approved treatment. Healx are pioneering the next generation of drug discovery by applying machine learning technology to accelerate the pace, increase the scale and improve the chance of success of rare disease treatment development. In this talk Dan O’Donovan and Dan Mason (both early employees and principal engineers / scientists in the research and development team) will discuss how the latest developments in computational chemistry, biology and machine learning research can benefit the lives of millions of patients.
Adrian Mulholland, University of Bristol
Multiscale Simulation for Drug Resistance and Enzyme Design
Molecular simulations are revealing mechanisms of drug resistance and allosteric effects in proteins, including the SARS-CoV-2 main protease and spike protein, and beta-lactamase enzymes that break down beta-lactam antibiotics. Dynamical-nonequilibrium molecular dynamics (D-NEMD) simulations are an emerging approach to identify allosteric communication pathways and distal positions associated with drug resistance mutations. Combined quantum mechanics/molecular mechanics (QM/MM) simulations reveal mechanisms of chemical reactions in proteins. They distinguish enzymes capable of breaking down specific antibiotics, and identify physical features that determine catalytic activity. Simulations can be used as ‘computational assays’ to predict functionally relevant properties, and contribute to de novo enzyme design.
For allosteric effects from D-NEMD simulations:
Oliveira, A.S.F., Ciccotti, G., Haider, S., Mulholland AJ. Dynamical nonequilibrium molecular dynamics reveals the structural basis for allostery and signal propagation in biomolecular systems. Eur. Phys. J. B 94, 144 (2021), https://link.springer.com/article/10.1140/epjb/s10051-021-00157-0
For enzyme design and evolution:
Adrian Bunzel, J.L. Ross Anderson, Adrian J. Mulholland, Designing better enzymes: Insights from directed evolution, Current Opinion in Structural Biology, 67, 2021, 212-218, https://www.sciencedirect.com/science/article/abs/pii/S0959440X21000075
For multiscale methods:
Amaro, R., Mulholland, A. Multiscale methods in drug design bridge chemical and biological complexity in the search for cures. Nat Rev Chem 2, 0148 (2018), https://doi.org/10.1038/s41570-018-0148, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6445369/
For interactive VR in molecular design:
Rebecca K. Walters, Ella M. Gale, Jonathan Barnoud, David R. Glowacki & Adrian J. Mulholland (2022) The emerging potential of interactive virtual reality in drug discovery, Expert Opinion on Drug Discovery, 17:7, 685-698, DOI: 10.1080/17460441.2022.2079632 , https://www.tandfonline.com/doi/full/10.1080/17460441.2022.2079632
Alexe Haywood, University of Nottingham
Kernel Methods for Predicting Yields of Chemical Reactions
The use of machine learning methods for the prediction of reaction yield is an emerging area. We demonstrate the applicability of support vector regression (SVR) for predicting reaction yields, using combinatorial data. Molecular descriptors used in regression tasks related to chemical reactivity have often been based on time-consuming, computationally demanding quantum chemical calculations, usually density functional theory. Structure-based descriptors (molecular fingerprints and molecular graphs) are quicker and easier to calculate, and are applicable to any molecule. In this study, SVR models built on structure-based descriptors were compared to models built on quantum chemical descriptors. The models were evaluated along the dimension of each reaction component in a set of Buchwald-Hartwig amination reactions (see below). The structure-based SVR models out-performed the quantum chemical SVR models, along the dimension of each reaction component. Prospective predictions of unseen Buchwald-Hartwig reactions are presented for synthetic assessment, to validate the generalisability of the models, with particular interest along the aryl halide dimension.T. Ahneman, J. G. Estrada, S. Lin, S. D. Dreher, A. G. Doyle, Science, 2018, 360, 186-190Alexe L. Haywood, Joseph Redshaw, Magnus W. D. Hanson-Heine, Adam Taylor, Alex Brown, Andrew M. Mason, Thomas Gärtner, and Jonathan D. Hirst, J. Chem. Inf. Model., 2022, 62, 9, 2077–2092.
Martin Slater, Cresset
Advancing molecular modelling projects through outsourcing