ChemistGA: A Chemical Synthesizable Accessible Molecular Generation Algorithm for Real-World Drug Discovery

Abstract

Many deep learning (DL)-based molecular generative models have been proposed to design novel molecules. These models may perform well on benchmarks, but they usually do not take real-world constraints into account, such as available training data set, synthetic accessibility, and scaffold diversity in drug discovery. In this study, a new algorithm, ChemistGA, was proposed by combining the traditional heuristic algorithm with DL, in which the crossover of the traditional genetic algorithm (GA) was redefined by DL in conjunction with GA, and an innovative backcrossing operation was implemented to generate desired molecules. Our results clearly show that ChemistGA not only retains the strength of the traditional GA but also greatly enhances the synthetic accessibility and success rate of the generated molecules with desired properties. Calculations on the two benchmarks illustrate that ChemistGA achieves impressive performance among the state-of-the-art baselines, and it opens a new avenue for the application of generative models to real-world drug discovery scenarios.

GENERA: A Combined Genetic/Deep-Learning Algorithm for Multiobjective Target-Oriented De Novo Design

“…GENERA proved to be a valuable tool for multiobjective optimization, as the generated focused libraries outperformed the starting reference library of ACE2 active compounds with regard to the objectives used during the generation. In summary, our algorithm quickly designed compounds (i) predicted to be even more affine toward the target than those in the starting reference set of known ACE-2 inhibitors, (ii) with good SA, which represents the main concern in de novo design projects, 50 and (iii) exploring a new chemical space. These results highlight GENERA’s potential as an innovative computational workflow for target-oriented de novo design, offering the flexibility to optimize, starting from a reference pool of compounds, relevant properties such as the predicted target affinity, drug-likeness, or any user-defined target-related property.…”

Augmenting Genetic Algorithms with Machine Learning for Inverse Molecular Design

“…By introducing ChemistGA (Figure 5), 93 Wang and co-workers demonstrated the use of an ANN-based crossover operation to account for synthetic accessibility during offspring generation in the evolutionary de novo design of drug-like compounds. The optimization goals were, in different combinations, protein activities for DRD2, JNK3, and GSK3β, druglikeness as measured by the QED score, 86 and synthetic accessibility as measured by the SA score.…”

LinChemIn: Route Arithmetic─Operations on Digital Synthetic Routes

“…The size of the corpus of chemical reaction data − obtained from public, proprietary, and licensed sources has progressively grown over the years, matched by an increasing demand for extracting more value from chemical experiments. , As a result, scientists are increasingly relying on computational tools for effective data navigation and analysis. For example, the Network of Organic Chemistry (NOC) , approach converts individual chemical reactions into a graph-like object, enabling powerful graph-based searches and facilitating the discovery of novel synthetic routes. , Additionally, advancements in predictive modeling has led to the development of Computer-Aided Synthesis Planning (CASP) tools, − which provide actionable insight in the form of synthetic plans to molecular targets. To address the need for common frameworks across multiple CASP/NOC tools, including the manual input of experts, we extended the Python toolkit LinChemIn with new functionalities.…”