Reinforced Adversarial Neural Computer for de Novo Molecular Design

By E. Putin, Arip Asadulaev, Y. Ivanenkov, V. Aladinskiy, Benjamín Sánchez-Lengeling, Alán Aspuru-Guzik, A. Zhavoronkov. Journal of Chemical Information and Modeling, 2018

325 citations recorded by the source corpus.

In silico modeling is a crucial milestone in modern drug design and development. Although computer-aided approaches in this field are well-studied, the application of deep learning methods in this research area is at the beginning. In this work, we present an original deep neural network (DNN) architecture named RANC (Reinforced Adversarial Neural Computer) for the de novo design of novel small-molecule organic structures based on the generative adversarial network (GAN) paradigm and reinforcement learning (RL). As a generator RANC uses a differentiable neural computer (DNC), a category of neural networks, with increased generation capabilities due to the addition of an explicit memory bank, which can mitigate common problems found in adversarial settings. The comparative results have shown that RANC trained on the SMILES string representation of the molecules outperforms its first DNN-based counterpart ORGANIC by several metrics relevant to drug discovery: the number of unique structures, passing medicinal chemistry filters (MCFs), Muegge criteria, and high QED scores. RANC is able to generate structures that match the distributions of the key chemical features/descriptors (e.g., MW, logP, TPSA) and lengths of the SMILES strings in the training data set. Therefore, RANC can be reasonably regarded as a promising starting point to develop novel molecules with activity against different biological targets or pathways. In addition, this approach allows scientists to save time and covers a broad chemical space populated with novel and diverse compounds.

View Paper: https://www.semanticscholar.org/paper/b4151c385bd30e352c18c0135981e56560ecf744

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

A Structure-Based Drug Discovery Paradigm (2019, methodology): “Relevant to drug discovery, RANC trained on SMILES string representation outperforms other methods on several metrics [124].”
Recent advances and applications of deep learning methods in materials science (2021, methodology): “There have also been several applications for identifying reasonable chemical space using DL methods such as autoenco-ders 111 and reinforcement learning 112 – 114 for inverse materials design.”
Potential of quantum computing for drug discovery (2018, background): “Similarly, state-of-the-art machine learning techniques, such as generative adversarial networks [64], are offering new paths towards de novo design of small molecules [65, 66].”
Self-referencing embedded strings (SELFIES): A 100% robust molecular string representation (2019, background): “41v1 [cs.LG] 31 May 2019 drugs and materials [4, 5]. Different generative models have been suggested for this task, such as variational autoencoders (VAEs) [6], generative adversarial networks (GANs) [7, 8, 9] or sampling from recurrent neural networks (RNNs) [10]. Objects of interest in the natural”
GuacaMol: Benchmarking Models for De Novo Molecular Design (2018, unknown): “…models based on neural networks have since been further explored, with work on VAEs, 17,36–45 RNNs, 46–58 generative adversarial networks (GANs), 54,55,59,60 adversarial au-toencoders (AAEs), 61–64 and graph convolutional networks, 16,57,60,65 using SMILES strings or graphs to represent molecules.”
All SMILES Variational Autoencoder (2019, methodology): “Reinforcement learning, often combined with adversarial methods, has been used to train progressive molecule growth strategies [17, 26, 47, 49, 70, 72].”