Quantifying the chemical beauty of drugs.

By G. Bickerton, G. V. Paolini, J. Besnard, Sorel Muresan, A. Hopkins. 2012

2,218 citations recorded by the source corpus.

Subfields: Computational Theory and Mathematics, 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

In Computational Chemistry & Drug Discovery

Citations in the corpus

Sequence Tutor: Conservative Fine-Tuning of Sequence Generation Models with KL-control (2016, background): “However, it should be noted that the Sequence Tutor model tends to produce simplistic molecules involving more carbon atoms than the MLE baseline; e.g. Sequence Tutor may produce ‘SNCc1ccccc1’, while the MLE produces ‘C(=O)c1ccc(S(=O)(=O)N(C)C)c(Cl)c1’, which is the reason for the Sequence Tutor mod”
Learning Multimodal Graph-to-Graph Translation for Molecular Optimization (2018, background): “Speciﬁcally, the model needs to translate molecules with QED scores (Bickerton et al., 2012) within the range [0 .”
Automatic Chemical Design Using a Data-Driven Continuous Representation of Molecules (2016, neutral): “37 f Shows the Qualitative Estimate of Drug-likeness (QED), 38 ranging from 0 to 1.”
Molecular de-novo design through deep reinforcement learning (2017, methodology): “They later extend this method to several other tasks including the generation of chemical structures, and optimize toward molecular properties such as cLogP [18] and QED drug-likeness [19].”
The ChEMBL bioactivity database: an update (2013, background): “…the well-used Lipinski Rule-of-Five (25); the Rule-of-Three passes, used to identify compounds suitable for fragment screening (26); and the weighted Quantitative Estimate of Drug likeness (QED_WEIGHTED), for which values range from 0–1 [1 being the most drug-like and 0 the least drug-like (27)].”
Structure-Based Virtual Screening for Drug Discovery: Principles, Applications and Recent Advances (2014, background): “Moreover, QED functions have the additional advantage that they are based on the underlying distribution data of drug properties and may identify cases in which a generally unfavorable property may be tolerated when the other parameters are close to ideal.”
Randomized SMILES strings improve the quality of molecular generative models (2019, methodology): “To prove that the molecules sampled from the randomized SMILES model are at least as diverse as those in the canonical, several physicochemical properties and metrics (as used in the MOSES benchmark [37]), such as molecular weight, logP, Synthetic Accessibility Score (SA) [38], Quantitative Estimate”
A Structure-Based Drug Discovery Paradigm (2019, methodology): “Sci. 2019, 20, 2783 13 of 18 DNN Deep neural network GAN Generative adversarial network ADMET Absorption, distribution, metabolism, excretion and toxicity GSK GlaxoSmithKline RANC Reinforced adversarial neural computer RL Reinforcement learning MD Molecular dynamics GPCRs G-protein-coupled receptors”
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].”
Fréchet ChemblNet Distance: A metric for generative models for molecules (2018, methodology): “Druglikeness [3] summarizes desired properties of molecules which are important e.”
MolGAN: An implicit generative model for small molecular graphs (2018, background): “The Quantitative Estimate of Druglikeness (QED) score quantiﬁes compound quality with a weighted geometric mean of desirability scores capturing the underlying data distribution of several drug properties (Bickerton et al., 2012).”
Computational Methods in Drug Discovery (2014, background): “Approximately 16% of oral drugs violate at least one criterion and 6% fail two or more criteria, and multiple examples exist of highly successful drugs that fail one or more of Lipinksi’s criteria including Lipitor and Singulair (Bickerton et al., 2012).”
Flow Network based Generative Models for Non-Iterative Diverse Candidate Generation (2021, background): “Note that for realistic drug design, we would need to consider many more quantities such as drug-likeness (Bickerton et al., 2012), toxicity, or synthesizability.”