Estimation of synthetic accessibility score of drug-like molecules based on molecular complexity and fragment contributions
By Peter Ertl, Ansgar Schuffenhauer. 2009
1,856 citations recorded by the source corpus.
BackgroundA method to estimate ease of synthesis (synthetic accessibility) of drug-like molecules is needed in many areas of the drug discovery process. The development and validation of such a method that is able to characterize molecule synthetic accessibility as a score between 1 (easy to make) and 10 (very difficult to make) is described in this article.ResultsThe method for estimation of the synthetic accessibility score (SAscore) described here is based on a combination of fragment contributions and a complexity penalty. Fragment contributions have been calculated based on the analysis of one million representative molecules from PubChem and therefore one can say that they capture historical synthetic knowledge stored in this database. The molecular complexity score takes into account the presence of non-standard structural features, such as large rings, non-standard ring fusions, stereocomplexity and molecule size. The method has been validated by comparing calculated SAscores with ease of synthesis as estimated by experienced medicinal chemists for a set of 40 molecules. The agreement between calculated and manually estimated synthetic accessibility is very good with r2 = 0.89.ConclusionA novel method to estimate synthetic accessibility of molecules has been developed. This method uses historical synthetic knowledge obtained by analyzing information from millions of already synthesized chemicals and considers also molecule complexity. The method is sufficiently fast and provides results consistent with estimation of ease of synthesis by experienced medicinal chemists. The calculated SAscore may be used to support various drug discovery processes where a large number of molecules needs to be ranked based on their synthetic accessibility, for example when purchasing samples for screening, selecting hits from high-throughput screening for follow-up, or ranking molecules generated by various de novo design approaches.
Subfields: Computational Theory and Mathematics, Materials Chemistry, Organic Chemistry, Computational Chemistry & Drug Discovery
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computational Theory and Mathematics
- José Miguel Hernández-Lobato
- Ola Engkvist
- Alán Aspuru-Guzik
- Esben Jannik Bjerrum
- Sergey I. Nikolenko
- Tianfan Fu
- Jimeng Sun
- Sepp Hochreiter
In Materials Chemistry
- José Miguel Hernández-Lobato
- Sergey I. Nikolenko
- Alán Aspuru-Guzik
- Sepp Hochreiter
- Shitong Luo
- Ola Engkvist
- Günter Klambauer
- Tianfan Fu
In Organic Chemistry
- Shitong Luo
- Alán Aspuru-Guzik
- Esben Jannik Bjerrum
- José Miguel Hernández-Lobato
- Richard Threlfall
- Thomas Kipf
- Nicola De Cao
- Olexandr Isayev
In Computational Chemistry & Drug Discovery
- Sepp Hochreiter
- Sergey I. Nikolenko
- Günter Klambauer
- Shitong Luo
- Philippe Schwaller
- José Miguel Hernández-Lobato
- Connor W. Coley
- Alán Aspuru-Guzik
Citations in the corpus
- Planning chemical syntheses with deep neural networks and symbolic AI (2017, background): “[26, 27] Second, the value of a position depends highly on the availability of suitable precursors.”
- Grammar Variational Autoencoder (2017, methodology): “As in Gómez-Bombarelli et al. (2016b) we consider a penalized logP score that takes into account other molecular properties such as ring size and synthetic accessibility (Ertl & Schuffenhauer, 2009).”
- Objective-Reinforced Generative Adversarial Networks (ORGAN) for Sequence Generation Models (2017, methodology): “• Synthetizability: A normalized version of the synthetic accessibility score [4] as implemented in RDKit, a measure based on molecular complexity and fragment contributions that estimates how hard (0) or how easy (1) it is to make a given molecule.”
- Introduction to the BioChemical Library (BCL): An Application-Based Open-Source Toolkit for Integrated Cheminformatics and Machine Learning in Computer-Aided Drug Discovery (2022, background): “Encouragingly, the molecules predicted most likely to be active at 100 nM (local PPV ≥ 0.80) have SAScores below 2.0, well-within an acceptable range (Ertl and Schuffenhauer, 2009).”
- Visualisation and subsets of the chemical universe database GDB-13 for virtual screening (2011, neutral): “perspective of synthetic and medicinal chemistry [31].”
- ChemTS: an efficient python library for de novo molecular generation (2017, methodology): “Following [11], we generate molecules that jointly optimize the octanol-water partition coefficient logP and the other twoproperties: synthetic accessibility [19] and ring penalty that penalizes unrealistically large rings.”
- Constrained Bayesian Optimization for Automatic Chemical Design (2017, background): “We assess molecular quality on the following objectives: m denotes a molecule, logP ( m ) is the water-octanol partition coefficient, QED ( m ) is the quantitative estimate of drug-likeness [Bickerton et al., 2012] and SA ( m ) is the synthetic accessibility [Ertl and Schuffenhauer, 2009].”
- Equivariant 3D-conditional diffusion model for molecular linker design (2022, positive): “2 Evaluation Metrics First, we report several chemical properties of the generated molecules that are especially important in drug design applications: average quantitative estimation of drug-likeness (QED) [4], average synthetic accessibility (SA) [12] and average number of rings in the linker.”