TOP: Towards Better Toxicity Prediction by Deep Molecular Representation Learning

Peng, Yuzhong; Zhang, Ziqiao; Jiang, Qizhi; Guan, Jihong; Zhou, Shuigeng

doi:10.1109/bibm47256.2019.8983340

Cited by 11 publications

(8 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Considering that chemical information is able to canonicalize SMILES strings, Peng et al [88] exploit a framework called TOP, combining the usage of SMILES and predefined properties, including root atom positions and isomeric features. A bidirectional gated recurrent unit-based RNN (BiGRU) is employed for the SMILES strings to capture local and global context information.…”

Section: Propertymentioning

confidence: 99%

“…As discovered in the molecule generation task, the overall performances on the MoleculeNet dataset reveal the overwhelming advantages of graph models over the string ones, especially for those involving multilevel structural details [109]. The superb performance of TOP [88] on the ClinTox dataset demonstrates the strengths of incorporating multiple kinds of fingerprints compared to SMILES2vec [86]. The speciallydesigned mixed representation with SMILES strings and physiochemical properties demonstrate its superior capability for toxicity prediction on the ClinTox dataset.…”

Section: Benchmark Analysismentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning Methods for Small Molecule Drug Discovery: A Survey

Hu¹,

Liu²,

Chen³

et al. 2023

Preprint

View full text Add to dashboard Cite

Section: Propertymentioning

confidence: 99%

Section: Benchmark Analysismentioning

confidence: 99%

Deep Learning Methods for Small Molecule Drug Discovery: A Survey

Hu¹,

Liu²,

Chen³

et al. 2023

Preprint

View full text Add to dashboard Cite

“…It is a benchmark toxicity prediction database that comprises 12 different cellular assays values of 8458 compounds corresponding to 12 different targets [19]. Twelves different cellular assays correspond to 12 classification tasks, although there are some missing values in each cellular assay [10], [19].…”

Section: A Benchmark Datasetsmentioning

confidence: 99%

“…Since QSPR can exploit the intrinsic relationships between molecular structure and the physicochemical and biochemical properties of molecules, QSPR models have been widely developed and applied to the ADMET properties prediction to provide faster, cost-effective and accurate prediction of unknown compounds ADMET properties [8], [9]. QSPR has become the most efficient and popular molecular properties prediction method due to its automatic, efficient, highthroughput, large-scale characteristics [10]. In QSPR models, multiple linear regression, neural network, random forest (RF), support vector machine, and fully connected-based deep neural network (FDNN, or DNN for short in some literature) are typically used to map the structure of compounds represented by hand-crafted features (i.e.…”

Section: Introductionmentioning

confidence: 99%

Enhanced Graph Isomorphism Network for Molecular ADMET Properties Prediction

et al. 2020

Self Cite

View full text Add to dashboard Cite

The evaluation of absorption, distribution, metabolism, exclusion, and toxicity (ADMET) properties plays a key role in a variety of domains including industrial chemicals, agrochemicals, cosmetics, environmental science, food chemistry, and particularly drug development. Since molecules are often intrinsically described as molecular graphs, graph neural networks have recently been studied to improve the prediction of ADMET properties. Among many graph neural networks published in recent years, Graph Isomorphism Network (GIN) is a relatively recent and very promising one. In this paper, we propose an enhanced GIN, called MolGIN, via exploiting the bond features and differences influence of the atom neighbors to end-to-end predict ADMET properties. Based on GIN, MolGIN concatenates the bond feature together with node feature in the feature aggregator and applies a gate unit to adjust the atomic neighborhood weights to map the differences in the interaction strength between the central atom and its neighbors, such that more meaningful structural patterns of molecules can be explored toward better molecular modeling. Extensive experiments were conducted on seven public datasets to evaluate MolGIN against four baseline models with benchmark metrics. Experimental results of MolGIN were also compared with state-of-the-art results published in the last three years on each dataset. Experimental results in terms of RMSE and AUC show that MolGIN significantly boosts the prediction performance of GIN and markedly outperforms the baseline models, and achieves comparable or superior performance to state-of-the-art results.

show abstract

“…74,[80][81][82][83][84][85][86][87] Convolutional neural networks (CNNs) and graph convolutional networks (GCNs) are two common architectures for discovering low-dimensional descriptors. CNNs employ convolutional layers to extract a low-dimensional set of spatial features present in a structured data set, like a pixel-or voxel-based digitized image shown in Figure3b.…”

mentioning

confidence: 99%

Machine learning-assisted design of material properties

Kadulkar,

Sherman,

Ganesan

et al. 2022

Preprint

View full text Add to dashboard Cite

Designing functional materials requires a deep search through multidimensional spaces for system parameters that yield desirable material properties. For cases where conventional parameter sweeps or trial-and-error sampling are impractical, inverse methods that frame design as a constrained optimization problem present an attractive alternative. However, even efficient algorithms require time-and resource-intensive characterization of material properties many times during optimization, imposing a design bottleneck. Approaches that incorporate machine learning can help address this limitation and accelerate the discovery of materials with targeted properties. Here, we review how to leverage machine learning to reduce dimensionality to effectively explore design space, accelerate property evaluation, and generate unconventional material structures with optimal properties. We also discuss promising future directions, including integration of machine learning into multiple stages of a design algorithm and interpretation of machine learning models to understand how design parameters relate to material properties.

show abstract

TOP: Towards Better Toxicity Prediction by Deep Molecular Representation Learning

Cited by 11 publications

References 22 publications

Deep Learning Methods for Small Molecule Drug Discovery: A Survey

Deep Learning Methods for Small Molecule Drug Discovery: A Survey

Enhanced Graph Isomorphism Network for Molecular ADMET Properties Prediction

Machine learning-assisted design of material properties

Contact Info

Product

Resources

About