Generative Models for Artificially‐intelligent Molecular Design

Automated De Novo Drug Design: Are We Nearly There Yet?

“…[80] Va rious instances of chemistry-savvy generative deep networks have already been developed for de novo drug design, with ac urrent emphasis on adversarial and reinforcement learning methods. [14,81] Theoretical studies are mushrooming,b ut only ahandful of prospective applications has been performed to date.C omputer scientists are well advised to develop algorithms that can detect meaningful patterns in small data sets,w hich are characteristic of earlystage drug discovery,a nd chemists should use these tools in prospective studies.H owever,c hemists and computer scientists often seem to be disconnected. While some of the theoreticians may consider the problem of de novo design solved in principle,w eo bserve am ix of enthusiasm, healthy scepticism, and even plain denial among medicinal chemists when it comes to de novo design.…”

“…Not surprisingly,t he advent of advanced machine learning techniques has also enabled the automated generation of new chemical entities with the desired properties. [14,15] In pioneering applications,t hese algorithms have demonstrated unique abilities to assemble molecules from basic building blocks (atoms,f ragments," tokens") while considering multiple properties and biological activities in parallel. Thee arly de novo design methods typically suffered from long runtimes and chemical ignorance.T oday,this situation has dramatically changed.…”

In silico Strategies to Support Fragment-to-Lead Optimization in Drug Discovery

“…More recently, a subfield of ML called deep learning (DL) which utilizes artificial neural networks to learn from a large amount of data have been used to resolve complex problems (Mak and Pichika, 2019). DL models are not only able to learn from a dataset and to make predictions for new data but are also able to generate new data instances through a constructive process (Schneider, 2018). In this context, there has been a rising interest in using DL generative and predictive models for F2L optimization (Olivecrona et al, 2017;Gupta et al, 2018).…”

Advances and challenges in deep generative models for de novo molecule generation

“…These molecules are supposed to come from the same training domain, having similar properties but distinct molecular structures. In practice, the labels y i in the training data are not necessary and the whole procedure can be trained in an unsupervised manner …”