DrugSynthMC: An Atom-Based Generation of Drug-like Molecules with Monte Carlo Search

Abstract

A growing number of deep learning (DL) methodologies have recently been developed to design novel compounds and expand the chemical space within virtual libraries. Most of these neural network approaches design molecules to specifically bind a target based on its structural information and/or knowledge of previously identified binders. Fewer attempts have been made to develop approaches for de novo design of virtual libraries, as synthesizability of generated molecules remains a challenge. In this work, we developed a new Monte Carlo Search (MCS) algorithm, DrugSynthMC (Drug Synthesis using Monte Carlo), in conjunction with DL and statistical-based priors to generate thousands of interpretable chemical structures and novel drug-like molecules per second. DrugSynthMC produces drug-like compounds using an atom-based search model that builds molecules as SMILES, character by character. Designed molecules follow Lipinski's "rule of 5″, show a high proportion of highly water-soluble nontoxic predicted-to-be synthesizable compounds, and efficiently expand the chemical space within the libraries, without reliance on training data sets, synthesizability metrics, or enforcing during SMILES generation. Our approach can function with or without an underlying neural network and is thus easily explainable and versatile. This ease in drug-like molecule generation allows for future integration of score functions aimed at different target- or job-oriented goals. Thus, DrugSynthMC is expected to enable the functional assessment of large compound libraries covering an extensive novel chemical space, overcoming the limitations of existing drug collections. The software is available at https://github.com/RoucairolMilo/DrugSynthMC.

Figure 1

AiZynthFinder synthetic routes search: (A) histogram plot showing the number of drugs from a data set comprising 909 drugs retrieved from the FDA subset of the ZINC20 database versus AiZynthFinder synthetic routes search time in seconds. (B) Histogram plot showing the number drugs for which AiZynthFinder successfully finds synthetic routes in less than 2 min versus drugs Molecular Weight. “Solved 1-step” drugs that can be produced directly from commercially available compounds; “Already in stock” drugs that are identified as commercially available by AiZynthFinder.

Figure 2

Physicochemical properties of generated drug-like Mmolecules. Comparison of physicochemical properties among drug-like compounds generated with Random and Ngram, and the “Rule of 5” drugs within the FDA subset of the ZINC20 database. Violin plots showing (A) number of hydrogen bond donors (HBD), (B) hydrogen bond acceptors (HBA), (C) molecular weight (MW), (D) total number of Carbons, (E) heavy atoms, (F) hydrogens, (G) nitrogen, (H) oxygen, (I) fluorine, (K) sulfur, (L) chlorine, (M) pentamers, (N) hexamers, (O) heptamers, (P) aromatic cycles, and (Q) clogP. Physicochemical properties were calculated on libraries containing 1,000 generated compounds.