ToxicBlend: Virtual screening of toxic compounds with ensemble predictors

Zaslavskiy, Mikhail; Jégou, Simon; Tramel, Eric W.; Wainrib, Gilles

doi:10.1016/j.comtox.2019.01.001

Cited by 24 publications

(23 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table 3 in Appendix C documents an even larger improvement compared to models that use a sparse Gaussian process for property prediction. We also surpass the state-of-the-art in toxicity prediction on the Tox21 dataset [22,45], as shown in Table 1, despite refraining from ensembling our model, or engineering features using expert chemistry knowledge, as in previous state-of-the-art methods [71].…”

Section: Property Predictionmentioning

confidence: 88%

See 1 more Smart Citation

All SMILES Variational Autoencoder

Alperstein,

Cherkasov,

Rolfe

2019

Preprint

View full text Add to dashboard Cite

Variational autoencoders (VAEs) defined over SMILES string and graph-based representations of molecules promise to improve the optimization of molecular properties, thereby revolutionizing the pharmaceuticals and materials industries. However, these VAEs are hindered by the non-unique nature of SMILES strings and the computational cost of graph convolutions. To efficiently pass messages along all paths through the molecular graph, we encode multiple SMILES strings of a single molecule using a set of stacked recurrent neural networks, pooling hidden representations of each atom between SMILES representations, and use attentional pooling to build a final fixed-length latent representation. By then decoding to a disjoint set of SMILES strings of the molecule, our All SMILES VAE learns an almost bijective mapping between molecules and latent representations near the high-probability-mass subspace of the prior. Our SMILES-derived but moleculebased latent representations significantly surpass the state-of-the-art in a variety of fully-and semi-supervised property regression and molecular property optimization tasks.Preprint. Under review.

show abstract

Section: Property Predictionmentioning

confidence: 88%

“…(a) ZINC250k MODEL MAE LOGP MAE QED ECFP [53] 0.38 0.045 CVAE [16] 0.15 0.054 CVAE ENC [16] 0.13 0.037 GRAPHCONV [12] [38] 0.854 POTENTIALNET [14] 0.857 ± 0.006 TOXICBLEND [71] 0.862 All SMILES 0.871…”

Section: Property Predictionmentioning

confidence: 99%

All SMILES Variational Autoencoder

Alperstein,

Cherkasov,

Rolfe

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Machine learning models to predict molecular properties have seen a large surge in popularity in the last decade, leading to new developments and impressive performances on the prediction of quantum-mechanical properties, [1][2][3] biological effects [4][5][6] or physicochemical properties, [7][8][9] to name just a few. In particular, graphbased approaches are on the rise, and have proven both powerful and useful in fields such as drug discovery.…”

Section: Introductionmentioning

confidence: 99%

Machine learning of reaction properties via learned representations of the condensed graph of reaction

Heid

Green

2021

Preprint

View full text Add to dashboard Cite

The estimation of chemical reaction properties such as activation energies, rates or yields is a central topic of computational chemistry. In contrast to molecular properties, where machine learning approaches such as graph convolutional neural networks (GCNNs) have excelled for a wide variety of tasks, no general and transferable adaptations of GCNNs for reactions have been developed yet. We therefore combined a popular cheminformatics reaction representation, the socalled condensed graph of reaction (CGR), with a recent GCNN architecture to arrive at a versatile, robust and compact deep learning model. The CGR is a superposition of the reactant and product graphs of a chemical reaction, and thus an ideal input for graph-based machine learning approaches. The model learns to create a data-driven, task dependent reaction embedding that does not rely on expert knowledge, similar to current molecular GCNNs. Our approach outperforms current stateof-the-art models in accuracy, is applicable even to imbalanced reactions and possesses excellent predictive capabilities for diverse target properties, such as activation energies, reaction enthalpies, rate constants, yields or reaction classes. We furthermore curated a large set of atom-mapped reactions along with their target properties, which can serve as benchmark datasets for future work. All datasets and the developed reaction GCNN model are available online, free of charge and opensource.

show abstract

“…Machine learning models to predict molecular properties have seen a large surge in popularity in the last decade, leading to new developments and impressive performances on the prediction of quantum-mechanical properties, 1-3 biological effects [4][5][6] or physicochemical properties, [7][8][9] to name just a few. In particular, graphbased approaches are on the rise, and have proven both powerful and useful in fields such as drug discovery.…”

Section: Introductionmentioning

confidence: 99%

Machine learning of reaction properties via learned representations of the condensed graph of reaction

Heid

Green

2021

Preprint

View full text Add to dashboard Cite

The estimation of chemical reaction properties such as activation energies, rates or yields is a central topic of computational chemistry. In contrast to molecular properties, where machine learning approaches such as graph convolutional neural networks (GCNNs) have excelled for a wide variety of tasks, no general and transferable adaptations of GCNNs for reactions have been developed yet. We therefore combined a popular cheminformatics reaction representation, the so-called condensed graph of reaction (CGR), with a recent GCNN architecture to arrive at a versatile, robust and compact deep learning model. The CGR is a superposition of the reactant and product graphs of a chemical reaction, and thus an ideal input for graph-based machine learning approaches. The model learns to create a data-driven, task dependent reaction embedding that does not rely on expert knowledge, similar to current molecular GCNNs. Our approach outperforms current state-of-the-art models in accuracy, is applicable even to imbalanced reactions and possesses excellent predictive capabilities for diverse target properties, such as activation energies, reaction enthalpies, rate constants, yields or reaction classes. We furthermore curated a large set of atom-mapped reactions along with their target properties, which can serve as benchmark datasets for future work. All datasets and the developed reaction GCNN model are available online, free of charge and open-source.

show abstract

ToxicBlend: Virtual screening of toxic compounds with ensemble predictors

Cited by 24 publications

References 32 publications

All SMILES Variational Autoencoder

All SMILES Variational Autoencoder

Machine learning of reaction properties via learned representations of the condensed graph of reaction

Machine learning of reaction properties via learned representations of the condensed graph of reaction

Contact Info

Product

Resources

About