TorsionNet: A Reinforcement Learning Approach to Sequential Conformer Search

Molecular geometry prediction of flexible molecules, or conformer search, is a long-standing challenge in computational chemistry. This task is of great importance for predicting structure-activity relationships for a wide variety of substances ranging from biomolecules to ubiquitous materials. Substantial computational resources are invested in Monte Carlo and Molecular Dynamics methods to generate diverse and representative conformer sets for medium to large molecules, which are yet intractable to chemoinformatic conformer search methods. We present TorsionNet, an efficient sequential conformer search technique based on reinforcement learning under the rigid rotor approximation. The model is trained via curriculum learning, whose theoretical benefit is explored in detail, to maximize a novel metric grounded in thermodynamics called the Gibbs Score. Our experimental results show that TorsionNet outperforms the highest scoring chemoinformatics method by 4x on large branched alkanes, and by several orders of magnitude on the previously unexplored biopolymer lignin, with applications in renewable energy.

[1]  Shie Mannor,et al.  Contextual Markov Decision Processes , 2015, ArXiv.

[2]  Andrew McCallum,et al.  Structured Prediction Energy Networks , 2015, ICML.

[3]  Peter Stone,et al.  Source Task Creation for Curriculum Learning , 2016, AAMAS.

[4]  Reid G. Simmons,et al.  Complexity Analysis of Real-Time Reinforcement Learning , 1993, AAAI.

[5]  Jivko Sinapov,et al.  Curriculum Learning for Reinforcement Learning Domains: A Framework and Survey , 2020, J. Mach. Learn. Res..

[6]  Mohammad M. Sultan,et al.  Variational encoding of complex dynamics. , 2017, Physical review. E.

[7]  Sereina Riniker,et al.  Better Informed Distance Geometry: Using What We Know To Improve Conformation Generation , 2015, J. Chem. Inf. Model..

[8]  Lignin-KMC: A Toolkit for Simulating Lignin Biosynthesis , 2019, ACS Sustainable Chemistry & Engineering.

[9]  Sanja Fidler,et al.  NerveNet: Learning Structured Policy with Graph Neural Networks , 2018, ICLR.

[10]  Charlotte M. Deane,et al.  Freely Available Conformer Generation Methods: How Good Are They? , 2012, J. Chem. Inf. Model..

[11]  Jason Weston,et al.  Curriculum learning , 2009, ICML '09.

[12]  G. P. Moss Basic terminology of stereochemistry (IUPAC Recommendations 1996) , 1996 .

[13]  Benjamin Lindner,et al.  Scaling of Multimillion-Atom Biological Molecular Dynamics Simulation on a Petascale Supercomputer. , 2009, Journal of chemical theory and computation.

[14]  Debora S. Marks,et al.  Learning Protein Structure with a Differentiable Simulator , 2018, ICLR.

[15]  Nando de Freitas,et al.  Reinforcement and Imitation Learning for Diverse Visuomotor Skills , 2018, Robotics: Science and Systems.

[16]  Marcin Andrychowicz,et al.  Solving Rubik's Cube with a Robot Hand , 2019, ArXiv.

[17]  Christof H. Schwab,et al.  Conformations and 3D pharmacophore searching. , 2010, Drug discovery today. Technologies.

[18]  B. Brooks,et al.  Self-guided Langevin dynamics simulation method , 2003 .

[19]  H. Kulik,et al.  Depolymerization Pathways for Branching Lignin Spirodienone Units Revealed with ab Initio Steered Molecular Dynamics. , 2017, The journal of physical chemistry. A.

[20]  P. Anastas,et al.  Designing for a green chemistry future , 2020, Science.

[21]  Anita R. Maguire,et al.  Confab - Systematic generation of diverse low-energy conformers , 2011, J. Cheminformatics.

[22]  Samy Bengio,et al.  Order Matters: Sequence to sequence for sets , 2015, ICLR.

[23]  Sergey Levine,et al.  Trust Region Policy Optimization , 2015, ICML.

[24]  Pieter Abbeel,et al.  Reverse Curriculum Generation for Reinforcement Learning , 2017, CoRL.

[25]  Thomas A. Halgren,et al.  Merck molecular force field. IV. conformational energies and geometries for MMFF94 , 1996, J. Comput. Chem..

[26]  Elman Mansimov,et al.  Molecular Geometry Prediction using a Deep Generative Graph Neural Network , 2019, Scientific Reports.

[27]  Sebastian Ruder,et al.  An overview of gradient descent optimization algorithms , 2016, Vestnik komp'iuternykh i informatsionnykh tekhnologii.

[28]  Garrett M Morris,et al.  Bayesian optimization for conformer generation , 2018, Journal of Cheminformatics.

[29]  Jure Leskovec,et al.  Graph Convolutional Policy Network for Goal-Directed Molecular Graph Generation , 2018, NeurIPS.

[30]  C. Levinthal How to fold graciously , 1969 .

[31]  Yanran Li,et al.  Adversarial Deep Reinforcement Learning in Portfolio Management , 2018 .

[32]  P. Hawkins Conformation Generation: The State of the Art , 2017, J. Chem. Inf. Model..

[33]  José Miguel Hernández-Lobato,et al.  A Generative Model for Molecular Distance Geometry , 2020, ICML.

[34]  Chris Morley,et al.  Open Babel: An open chemical toolbox , 2011, J. Cheminformatics.

[35]  C. Bolm,et al.  Mechanochemical degradation of lignin and wood by solvent-free grinding in a reactive medium , 2013 .

[36]  Koji Tsuda,et al.  ChemTS: an efficient python library for de novo molecular generation , 2017, Science and technology of advanced materials.

[37]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[38]  Ambuj Tewari,et al.  No-regret Exploration in Contextual Reinforcement Learning , 2019, UAI.

[39]  Demis Hassabis,et al.  Improved protein structure prediction using potentials from deep learning , 2020, Nature.

[40]  A. C. Buchanan,et al.  Computational investigation of the pyrolysis product selectivity for α-hydroxy phenethyl phenyl ether and phenethyl phenyl ether: analysis of substituent effects and reactant conformer selection. , 2013, The journal of physical chemistry. A.

[41]  Jianpeng Ma,et al.  CHARMM: The biomolecular simulation program , 2009, J. Comput. Chem..

[42]  Katalin Barta,et al.  Bright Side of Lignin Depolymerization: Toward New Platform Chemicals , 2018, Chemical reviews.

[43]  Mike Preuss,et al.  Planning chemical syntheses with deep neural networks and symbolic AI , 2017, Nature.

[44]  Matthias Rarey,et al.  TFD: Torsion Fingerprints As a New Measure To Compare Small Molecule Conformations , 2012, J. Chem. Inf. Model..

[45]  Michael Gastegger,et al.  Generating equilibrium molecules with deep neural networks , 2018, ArXiv.

[46]  Jan Eric Lenssen,et al.  Fast Graph Representation Learning with PyTorch Geometric , 2019, ArXiv.

[47]  Zheng Wen,et al.  Efficient Exploration and Value Function Generalization in Deterministic Systems , 2013, NIPS.

[48]  Gerald A. Tuskan,et al.  Lignin Valorization: Improving Lignin Processing in the Biorefinery , 2014, Science.

[49]  Alexander D. MacKerell,et al.  CHARMM general force field: A force field for drug‐like molecules compatible with the CHARMM all‐atom additive biological force fields , 2009, J. Comput. Chem..

[50]  Samuel S. Schoenholz,et al.  Neural Message Passing for Quantum Chemistry , 2017, ICML.

[51]  D. Weinshall,et al.  Curriculum Learning by Transfer Learning: Theory and Experiments with Deep Networks , 2018, ICML.

[52]  Wenli Song,et al.  Initial Mechanisms for an Overall Behavior of Lignin Pyrolysis through Large-Scale ReaxFF Molecular Dynamics Simulations , 2016 .

[53]  Florian Sittel,et al.  Principal component analysis on a torus: Theory and application to protein dynamics. , 2017, The Journal of chemical physics.