Automated De Novo Design in Medicinal Chemistry: Which Types of Chemistry Does a Generative Neural Network Learn?

Abstract

Artificial intelligence offers promising solutions for property prediction, compound design, and retrosynthetic planning, which are expected to significantly accelerate the search for pharmacologically relevant molecules. Here, we investigate aspects of artificial intelligence based de novo design pertaining to its integration into real-life workflows. First, different chemical spaces were used as training sets for reinforcement learning (RL) in combination with different reward functions. With the trained neuronal networks different biologically active molecules could be regenerated. Excluding molecules with substructures such as five-membered rings from training spaces nevertheless produced results containing these moieties. Furthermore, different scoring functions in RL were investigated and produced different design ensembles. In summary, some of these design proposals are close in chemical space to the query, thus supporting lead optimization, while 3D-shape or QSAR (quantitative structure−activity relationship) models produced significantly different proposals by sampling a broader region of the chemical space, thus supporting lead generation. Therefore, RL provides a good framework to tailored design approaches for different discovery phases.

LibINVENT: Reaction-based Generative Scaffold Decoration for in Silico Library Design

“…The rewards are shaped by a scoring function defining desirable chemical or structural properties and guide the model towards producing compounds of interest 8 . However, since the objective is to explore a rather narrow space of solutions (molecules) designed for a given scaffold, this may lead to an mode collapse 30 .…”

“…One of these specific requirements is in the lead optimization stage when aiming to use focused libraries of small molecules to identify a promising lead compound 7,8 . Generally, the purpose of lead optimization is to retain the favourable properties of the compound while optimizing properties which still prevents the compound from becoming a drug candidate 9 .…”

Allosteric Regulation at the Crossroads of New Technologies: Multiscale Modeling, Networks, and Machine Learning

“…After years of serving as a mere inspiration rather than a practical tool, DRL techniques have taken off overcoming scalability and data limitation issues, and exploding into one of the most intense areas of AI research. Recent years have witnessed the expansion of DRL applications into biomedical research including but not limited to biomedical informatics, drug discovery (Baskin, 2020 ; Grebner et al, 2020 ), and toxicology (Chary et al, 2020 ).…”

REINVENT 2.0 – an AI Tool for De Novo Drug Design

“…To address such needs the research community has recently turned its focus towards artificial intelligence (AI) based generative models that are capable of proposing promising small molecules. The potential of generative models for chemical space exploration has been demonstrated in numerous studies [1][2][3][4][5][6][7][8][9][10][11][12][13] . Various neural network architectures have been engineered and a plethora of AI training strategies have been employed in the race to device more efficient methods for the generation of compounds.…”