Prospective evaluation of structure-based simulations reveal their ability to predict the impact of kinase mutations on inhibitor binding

Sukrit Singh, Vytautas Gapsys, Matteo Aldeghi, David Schaller, Aziz M Rangwala, Jessica B White, Joseph P Bluck, Jenke Scheen, William G Glass, Jiaye Guo, Sikander Hayat, Bert L de Groot, Andrea Volkamer, Clara D Christ, Markus A Seeliger, John D Chodera.
[bioRxiv]

We show that alchemical free energy calculations have the potential to prospectively predict the impact of clinical kinase mutations on targeted kinase inhibitor binding.

Nutmeg and SPICE: Models and data for biomolecular machine learning

Peter Eastman, Benjamin P. Pritchard, John D. Chodera, Thomas E. Markland
Journal of Chemical Theory and Computation 20:8583, 2024.
[DOI] [preprint]

We present a significant expansion of the SPICE dataset, a large-scale quantum chemical dataset for training machine learning potentials, and show how it can be used to build extremely accurate machine learning potentials.

Lessons learned during the journey of data: from experiment to model for predicting kinase affinity, selectivity, polypharmacology, and resistance

Raquel López-Ríos de Castro, Jaime Rodríguez-Guerra, David Schaller, Talia B Kimber, Corey Taylor, Jessica B White, Michael Backenköhler, Alexander Payne, Ben Kaminow, Iván Pulido, Sukrit Singh, Paula Linh Kramer, Guillermo Pérez-Hernández, Andrea Volkamer, John D Chodera
[bioRxiv]

This best practices paper describes considerations relevant to the use of experimental datasets in structure-based machine learning, using kinase:small molecule interactions as a model system.

Machine-learned molecular mechanics force fields from large-scale quantum chemical data

Kenichiro Takaba, Anika J Friedman, Chapin E Cavender, Pavan Kumar Behara, Iván Pulido, Michael M Henry, Hugo MacDermott-Opeskin, Christopher R Iacovella, Arnav M Nagle, Alexander Matthew Payne, Michael R Shirts, David L Mobley, John D Chodera, Yuanqing Wang
Chemical Science 15:12861, 2024 [DOI] [arXiv preprint]

We present a new self-consistent MM force field trained on $>$1.1M quantum chemical calculations that uses graph nets to achieve high accuracy and produce accurate protein-ligand binding free energies.

DrugGym: A testbed for the economics of autonomous drug discovery

Michael Retchin, Yuanqing Wang, Kenichiro Takaba, and John D. Chodera
[bioRxiv preprint]

We present DrugGym, a sandbox for exploring reinforcement learning strategies and evaluating the economics of decisionmaking strategies and predictive models on small molecule discovery. We use this tool to quantify the value of predictive model accuracy on hit-to-lead programs.

Enhancing protein–ligand binding affinity predictions using neural network potentials

Francesc Sabanés Zariquiey, Raimondas Galvelis, Emilio Gallicchio, John D. Chodera, Thomas E. Markland, Gianni De Fabritiis
Journal of Chemical Information and Modeling 64:1481, 2024.
[DOI] [preprint]

We show that hybrid neural network / molecular mechanics potentials can significantly improve accuracy over molecular mechanics potentials alone in predicting protein-ligand binding affinities.

Death by a thousand cuts through kinase inhibitor combinations that maximize selectivity and enable rational multitargeting

Outhwaite IR, Singh S, Berger B-T, Knapp S, Chodera JD, Seeliger MA
eLife 12:e86189, 2024 [DOI] [bioRxiv] [GitHub]

We show how combinations of kinase inhibitors can achieve selectivity gains for rational kinase polypharmacology.

OpenMM 8: Molecular Dynamics Simulation with Machine Learning Potentials

Eastman P, Galvelis R, Peláez RP, Abreu CRA, Farr SE, Gallicchio E, Gorenko A, Henry MH, Hu F, Huang J, Krämer A, Michel J, Mitchell J, Pande VS, Rodrigues JPGLM, Rodriguez-Guerra J, Simmonett AC, Swails J, Turner P, Wang Y, Zhang I, Chodera JD, De Fabritiis G, Markland TE
Journal of Physical Chemistry B [DOI] [website] [code]

We present OpenMM 8, which includes GPU-accelerated support for simulating hybrid ML/MM systems that use machine learning (ML) potentials to achieve high accuracy with minimal loss in speed.

Open science discovery of potent noncovalent SARS-CoV-2 main protease inhibitors

Boby ML, Fearon D, Ferla M, Filep M, Robinson MC, The COVID Moonshot Consortium, Chodera JD, Lee A, London N, von Delft F.
Science 382:eabo7201, 2023 [DOI] [PDF] [ready to use data]

We report the discovery of a new oral antiviral non-covalent SARS-CoV-2 main protease inhibitor developed by the COVID Moonshot, a global open science collaboration leveraging free energy calculations on Folding@home and ML-accelerated synthesis planning, now in accelerated preclinical studies funded by an $11M grant from the WHO ACT-A program via the Wellcome Trust. We are currently in discussions with generics manufacturers about partnering with us throughout clinical trials to ensure we can scale up production for global equitable and affordable access once approved by regulatory agencies.

Benchmarking cross-docking strategies for structure-informed machine learning in kinase drug discovery

Schaller D, Christ CD, Chodera JD, Volkamer A
preprint: [bioRxiv]

We assess strategies for predicting useful docked ligand poses for structure-informed machine learning for kinase inhibitor drug discovery.

NNP/MM: Fast molecular dynamics simulations with machine learning potentials and molecular mechanics

Galvelis R, Varela-Rial A, Doerr S, Fino R, Eastman P, Markland TE,  Chodera JD, and de Fabritiis G
Journal of Chemical Information and Modeling 63:5701, 2023 [DOI] [arXiv]

We demonstrate that a new generation of quantum machine learning (QML) potentials based on neural networks---which can achieve quantum chemical accuracy at a fraction of the cost---can be implemented efficiently in the OpenMM molecular dynamics simulation engine as part of hybrid machine learning / molecular mechanics (ML/MM) potentials that promise to deliver superior accuracy for modeling protein-ligand interactions.

Identifying and Overcoming the Sampling Challenges in Relative Binding Free Energy Calculations of a Model Protein:Protein Complex

Zhang I, Rufa DA, Pulido I, Henry MM, Rosen LE, Hauser K, Singh S, Chodera JD
Journal of Chemical Theory and Computation 19:4863, 2023

We assess what is required for alchemical free energy calculations to be able to make high-quality predictions of the impact of interfacial mutations on protein-protein binding.

Development and benchmarking of Open Force Field 2.0.0---the Sage small molecule force field

Boothroyd S, Behara PK, Madin OC, Hahn DF, Jang H, Gapsys V, Wagner JR, Horton JT, Dotson DL, Thompson MW, Maat J, Gokey T, Wang L-P, Cole DJ, Gilson MK, Chodera JD, Bayly CI, Shirts MR, Mobley DL
Journal of Chemical Theory and Computation 19:3251, 2023 [DOI] [chemRxiv] [GitHub] [examples]

We present a new generation of small molecule force field for molecular design from the Open Force Field Initiative fit to both quantum chemical and experimental liquid mixture data

MEN1 mutations mediate clinical resistance to menin inhibition

Perner F, Stein EM, Wenge DV, Singh S, Kim J, Apazidis A, Rahnamoun H, Anand D, Marinaccio C, Hatton C, Wen Y, Stone RM, Schaller D, Mowla S, Xiao W, Gamlen HA, Stonestrom AJ, Persaud S, Ener E, Cutler JA, Doench JG, McGeehan GM, Volkamer A, Chodera JD, Nowak RP, Fischer ES, Levine RL, Armstrong SA, Cai SF
Nature 615:913, 2023 [DOI]

We describe how mutants that confer therapeutic resistance to menin inhibition impact small molecule binding but not interactions with the natural ligand MLL1.

Turning high-throughput structural biology into predictive inhibitor design

Saar KL, McCorkindale W, Fearon D, Boby M, Barr H, Ben-Shmuel A, COVID Moonshot Consortium, London N, von Delft F, Chodera JD, Lee AA
PNAS 120:e2214168120, 2023 [DOI]

We demonstrate how potent inhibitors can be predicted from high-throughput structural biology, demonstrating this approach against the SARS-CoV-2 main viral protease (Mpro).

EspalomaCharge: Machine learning-enabled ultra-fast partial charge assignment

Wang Y, Pulido I, Takaba K, Kaminow B, Scheen J, Wang L, Chodera JD
preprint: [arXiv]

We present a drop-in replacement for generating AM1-BCC ELF10 charges based on graph convolutional nets that is orders of magnitude faster than standard methods for both small molecules and biomolecules.

Spatial attention kinetic network with E(n) equivariance

Yuanqing Wang and John D. Chodera
preprint: [arXiv] [code]

This work descibes Spatial Attention Kinetic Networks (SAKE), a new E(n)-equivariant architecture that uses spatial attention, enabling the construction of extremely performant but still accurate machine learning potentials, as well as flows capable of prediction dynamics.

SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine Learning Potentials

Eastman P, Behara PK, Dotson DL, Galvelis R, Herr JE, Horton JT, Mao Y, Chodera JD, Pritchard BP, Wang Y, De Fabritiis G, and Markland TE
Scientific Data 10:11, 2023 [DOI]

To remedy the lack of large, open quantum chemical datasets for training accurate general machine learning potentials and molecular mechanics force fields for druglike small molecules and biomolecules, we produce the open SPICE dataset, and show how it can be used to build extremely accurate machine learning potentials.

Open Force Field BespokeFit: Automating Bespoke Torsion Parametrization at Scale

Horton JT, Boothroyd S, Wagner W, Mitchell JA, Gokey T, Dotson DL, Behara PK, Ramaswamy VK, Mackey M, Chodera JD, Anwar J, Mobley DL, and Cole DJ
Journal of Chemical Informatics and Modeling 62:22, 2022 [DOI]

We describe an automated pipeline for generating tailored force field parameters for small molecules using quantum chemical or quantum machine learning potentials.

End-to-end differentiable molecular mechanics force field construction

Yuanqing Wang, Josh Fass, and John D. Chodera
Chemical Science 13:12016, 2022 [DOI] [arXiv] [pytorch code] [JAX code]

Molecular mechanics force fields have been a workhorse for computational chemistry and drug discovery. Here, we propose a new approach to force field parameterization in which graph convolutional networks are used to perceive chemical environments and assign molecular mechanics (MM) force field parameters. The entire process of chemical perception and parameter assignment is differentiable end-to-end with respect to model parameters, allowing new force fields to be easily constructed from MM or QM force fields, extended, and applied to arbitrary biomolecules.