Deep Reinforcement Learning and Docking Simulations for autonomous molecule generation in de novo Drug Design

Due to the large drug-like chemical space available to search for feasible drug-like molecules, rational drug design often starts from specific scaffolds to which side chains/substituents are added or modified. With the rapid growth of the application of deep learning in drug discovery, a variety of effective approaches have been developed for de novo drug design. In previous work, we proposed a method named DrugEx, which can be applied in polypharmacology based on multi-objective deep reinforcement learning. However, the previous version is trained under fixed objectives similar to other known methods and does not allow users to input any prior information (i.e. a desired scaffold). In order to improve the general applicability, we updated DrugEx to design drug molecules based on scaffolds which consist of multiple fragments provided by users. In this work, the Transformer model was employed to generate molecular structures. The Transformer is a multi-head self-attention deep learning model containing an encoder to receive scaffolds as input and a decoder to generate molecules as output. In order to deal with the graph representation of molecules we proposed a novel positional encoding for each atom and bond based on an adjacency matrix to extend the architecture of the Transformer. Each molecule was generated by growing and connecting procedures for the fragments in the given scaffold that were unified into one model. Moreover, we trained this generator under a reinforcement learning framework to increase the number of desired ligands. As a proof of concept, our proposed method was applied to design ligands for the adenosine A2A receptor (A2AAR) and compared with SMILES-based methods. The results demonstrated the effectiveness of our method in that 100% of the generated molecules are valid and most of them had a high predicted affinity value towards A2AAR with given scaffolds.

Download Full-text

Deep reinforcement learning for de novo drug design

Science Advances ◽

10.1126/sciadv.aap7885 ◽

2018 ◽

Vol 4 (7) ◽

pp. eaap7885 ◽

Cited By ~ 194

Author(s):

Mariya Popova ◽

Olexandr Isayev ◽

Alexander Tropsha

Keyword(s):

Reinforcement Learning ◽

Drug Design ◽

De Novo ◽

De Novo Drug Design

Download Full-text

Deep Reinforcement Learning for Multiparameter Optimization in de novo Drug Design

Journal of Chemical Information and Modeling ◽

10.1021/acs.jcim.9b00325 ◽

2019 ◽

Vol 59 (7) ◽

pp. 3166-3176 ◽

Cited By ~ 19

Author(s):

Niclas Ståhl ◽

Göran Falkman ◽

Alexander Karlsson ◽

Gunnar Mathiason ◽

Jonas Boström

Keyword(s):

Reinforcement Learning ◽

Drug Design ◽

De Novo ◽

De Novo Drug Design ◽

Multiparameter Optimization

Download Full-text

De novo drug design using reinforcement learning with graph-based deep generative models

10.33774/chemrxiv-2021-9w3tc ◽

2021 ◽

Author(s):

Sara Romeo Atance ◽

Juan Viguera Diez ◽

Ola Engkvist ◽

Simon Olsson ◽

Rocío Mercado

Keyword(s):

Reinforcement Learning ◽

Drug Design ◽

De Novo ◽

Generative Models ◽

De Novo Drug Design

Download Full-text

Advances in De Novo Drug Design: From Conventional to Machine Learning Methods

International Journal of Molecular Sciences ◽

10.3390/ijms22041676 ◽

2021 ◽

Vol 22 (4) ◽

pp. 1676

Author(s):

Varnavas D. Mouchlis ◽

Antreas Afantitis ◽

Angela Serra ◽

Michele Fratello ◽

Anastasios G. Papadiamantis ◽

...

Keyword(s):

Machine Learning ◽

Neural Networks ◽

Reinforcement Learning ◽

Drug Design ◽

De Novo ◽

A Priori ◽

Building Blocks ◽

Molecular Structures ◽

Generative Adversarial Networks ◽

De Novo Drug Design

De novo drug design is a computational approach that generates novel molecular structures from atomic building blocks with no a priori relationships. Conventional methods include structure-based and ligand-based design, which depend on the properties of the active site of a biological target or its known active binders, respectively. Artificial intelligence, including ma-chine learning, is an emerging field that has positively impacted the drug discovery process. Deep reinforcement learning is a subdivision of machine learning that combines artificial neural networks with reinforcement-learning architectures. This method has successfully been em-ployed to develop novel de novo drug design approaches using a variety of artificial networks including recurrent neural networks, convolutional neural networks, generative adversarial networks, and autoencoders. This review article summarizes advances in de novo drug design, from conventional growth algorithms to advanced machine-learning methodologies and high-lights hot topics for further development.

Download Full-text

Deep Reinforcement Learning for Multiparameter Optimization in de novo Drug Design

10.26434/chemrxiv.7990910.v2 ◽

2019 ◽

Author(s):

Niclas Ståhl ◽

Göran Falkman ◽

Alexander Karlsson ◽

Gunnar Mathiason ◽

Jonas Boström

Keyword(s):

Reinforcement Learning ◽

Short Term Memory ◽

De Novo ◽

De Novo Drug Design ◽

Generative Process ◽

New Methods ◽

Multiparameter Optimization ◽

Long Short Term Memory ◽

New Compounds

<p>In medicinal chemistry programs it is key to design and make compounds that are efficacious and safe. This is a long, complex and difficult multi-parameter optimization process, often including several properties with orthogonal trends. New methods for the automated design of compounds against profiles of multiple properties are thus of great value. Here we present a fragment-based reinforcement learning approach based on an actor-critic model, for the generation of novel molecules with optimal properties. The actor and the critic are both modelled with bidirectional long short-term memory (LSTM) networks. The AI method learns how to generate new compounds with desired properties by starting from an initial set of lead molecules and then improve these by replacing some of their fragments. A balanced binary tree based on the similarity of fragments is used in the generative process to bias the output towards structurally similar molecules. The method is demonstrated by a case study showing that 93% of the generated molecules are chemically valid, and a third satisfy the targeted objectives, while there were none in the initial set.</p>

Download Full-text

Lead Molecules as Novel Aromatase Inhibitors: In Silico De Novo Designing and Binding Affinity Studies

Letters in Drug Design & Discovery ◽

10.2174/1570180816666190703152659 ◽

2020 ◽

Vol 17 (5) ◽

pp. 655-665 ◽

Cited By ~ 1

Author(s):

Laxmi Banjare ◽

Sant Kumar Verma ◽

Akhlesh Kumar Jain ◽

Suresh Thareja

Keyword(s):

Drug Design ◽

Aromatase Inhibitors ◽

De Novo ◽

Scale Up ◽

Geometry Optimization ◽

Docking Study ◽

Molecular Docking Study ◽

De Novo Drug Design ◽

Molegro Virtual Docker ◽

Estrogen Production

Background:Aromatase inhibitors emerged as a pivotal moiety to selectively block estrogen production, prevention and treatment of tumour growth in breast cancer. De novo drug design is an alternative approach to blind virtual screening for successful designing of the novel molecule against various therapeutic targets.Objective:In the present study, we have explored the de novo approach to design novel aromatase inhibitors.Method:The e-LEA3D, a computational-aided drug design web server was used to design novel drug-like candidates against the target aromatase. For drug-likeness ADME parameters (molecular weight, H-bond acceptors, H-bond donors, LogP and number of rotatable bonds) of designed molecules were calculated in TSAR software package, geometry optimization and energy minimization was accomplished using Chem Office. Further, molecular docking study was performed in Molegro Virtual Docker (MVD).Results:Among 17 generated molecules using the de novo pathway, 13 molecules passed the Lipinski filter pertaining to their bioavailability characteristics. De novo designed molecules with drug-likeness were further docked into the mapped active site of aromatase to scale up their affinity and binding fitness with the target. Among de novo fabricated drug like candidates (1-13), two molecules (5, 6) exhibited higher affinity with aromatase in terms of MolDock score (-150.650, -172.680 Kcal/mol, respectively) while molecule 8 showed lowest target affinity (-85.588 Kcal/mol).Conclusion:The binding patterns of lead molecules (5, 6) could be used as a pharmacophore for medicinal chemists to explore these molecules for their aromatase inhibitory potential.

Download Full-text