enchmark for molecular machine learning †

By Zhenqin Wu, Bharath Ramsundar, Evan N. Feinberg, Joseph Gomes, Caleb Geniesse, Aneesh S. Pappu, Karl Leswingd, Vijay S. Pande. 2017

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Xiang Wang
2. Xiangnan He
3. Tat-seng Chua
4. Kyle Swanson
5. Fei Wang
6. Chen Gao
7. Kevin Yang
8. Regina Barzilay

Citations in the corpus

• Ligand biological activity predicted by cleaning positive and negative chemical correlations (2019, methodology): “Table 1 also shows that RMD outperforms SVM as well as the Graph Convolutions [implemented in the open source package DeepChem (34)].”
• AGL-Score: Algebraic Graph Learning Score for Protein-Ligand Binding Scoring, Ranking, Docking, and Screening (2019, methodology): “Graph theory has also been used to represent the structures of molecules and biomolecules, resulting in a popular approach for chemical data-sets 2,10,11,24 − 26 and biomolecular datasets 3,27 − 32 in the past few decades.”
• Enhanced Deep‐Learning Prediction of Molecular Properties via Augmentation of Bond Topology (2018, methodology): “9 GCN for multidisciplinary tasks and showed good performance regardless of the dataset categories.(30,39) Our cross-validation results showed that the 3DGCNsum had consistently better performance than the Weave model for all datasets.”
• Graph convolutional networks for computational drug development and discovery (2019, background): “Tox21 [57],ToxCast [87] and ClinTox [63] are datasets containing molecular toxicity information.”
• ToxicBlend: Virtual Screening of Toxic Compounds with Ensemble Predictors (2018, methodology): “We tested every molecular feature subset in conjunction with the a few machine learning models trained in a multi-task setting, namely: extreme gradient boosting trees (Xgboost) [35], DNNs [36], as well as GCNs as implemented in the DeepChem package [24].”
• Uncertainty Quantification Using Neural Networks for Molecular Property Prediction (2020, methodology): “18 We systematically evaluate a variety of UQ techniques compatible with NNs. 23 In addition, we also apply these methods to feed forward networks (FFNs) using static fingerprint (FP) representations, to see if behavior is constant across NN architectures.”