DeepChem Papers and Discoveries List

Paper Title: Drugs repurposing using QSAR, docking and molecular dynamics for possible inhibitors of the SARS-CoV-2 Mpro protease
Summary of DeepChem Usage: DeepChem was used extensively to train models. Code is available at
Important Contributions: ML methods were used to investigate potential repurposed compounds for covid targeting Mpro
Date Published: 2020
Journal: Molecules

Paper Title: Improvement of Prediction Performance With Conjoint Molecular Fingerprint in Deep Learning
Summary of DeepChem Usage: The scaffold splitter, and ECFP fingerprints from DeepChem were used.
Important Contributions: A joint featurization method combining two standard molecular fingerprints is investigated.
Date Published: December 18, 2018
Journal: Frontiers in Pharmacology

Paper Title: A Self-Improving Photosensitizer Discovery System via Bayesian Optimization and Quantum Chemical Calculation
Summary of DeepChem Usage: Surrogate models were trained using DeepChem graphconvs.
Important Contributions: Using bayesian optimization plus graph convolutions, a high performance sensitizer is discovered.
Date Published: 2021
Journal: Chemrxiv

Paper Title: Investigating the Application of Interpretability Techniques to Computational Toxicology
Summary of DeepChem Usage: DeepChem is used extensively for datasets and modeling.
Important Contributions: Interpretabillity methods in computational toxicology are explored.
Date Published: 2021
Journal: UMD Honors Thesis

1 Like

Paper Title: Analyzing learned molecular representations for property prediction
Summary of DeepChem Usage: DeepChem/MoleculeNet models datasets are used to benchmark the new model.
Important Contributions: The D-MPNN model is one of the leading ML for chemistry models invented in the last couple of years.
Date Published: July 30, 2019
Journal: Journal of Chemical Informatics and Modeling

Paper Title: Pushing the boundaries of molecular representation for drug discovery with the graph attention mechanism
Summary of DeepChem Usage: DeepChem GraphConvs and MPNNs were used to baseline the new algorithms in this work.
Important Contributions: The new AttentiveFP model is introduced.
Date Published: August 13, 2019
Journal: Journal of Medicinal Chemistry

Paper Title: Potential neutralizing antibodies discovered for novel corona virus using machine learning
Summary of DeepChem Usage: DeepChem code is not directly used but the featurization code is adapted from DeepChem featurization code (compare,
Important Contributions: Machine learning is used to identify potential neutralizing antibodies for covid using ML.
Date Published: March 4th, 2021
Journal: Nature Scientific Reports

Paper Title: GraphDTA: prediction of drug–target binding affinity using graph convolutional networks
Summary of DeepChem Usage: Atomic features for graphs are adapted from deepchem.
Important Contributions: Graph convolutional networks are used to predict drug/target interactions.
Date Published: October 2nd, 2020
Journal: Biorxiv

Paper Title: Padme: A deep learning-based framework for drug-target interaction prediction
Summary of DeepChem Usage: Weave featurizer and model adapted from DeepChem for their framework.
Important Contributions: PADME combines graph convolutions plus protein descriptors to make a protein-ligand interaction strength predictor.
Date Published: August 21st, 2019
Journal: Arxiv

Paper Title: Molecule attention transformer
Summary of DeepChem Usage: SVM, RF, Weave, and GraphConv models were used as baselines (from DeepCHem)
Important Contributions: Introduces the molecule attention transformer architecture.
Date Published: February 19th, 2020
Journal: Arxiv

Paper Title: Modeling physico-chemical ADMET endpoints with multitask graph convolutional networks
Summary of DeepChem Usage: Multitask graph convolutional networks are used directly.
Important Contributions: This paper generates that graph-convs are a powerful tool for modelling physico-chemical ADMET endpoints.
Date Published: December 1st, 2019
Journal: Molecules

Paper Title: Predicting or pretending: artificial intelligence for protein-ligand interactions lack of sufficiently large and unbiased datasets
Summary of DeepChem Usage: Atomic convolutions are used to train a protein ligand model from deepchem.
Important Contribution: Data biases in PDBBind and DUD-E are investigated using ACNNs.
Date Published: February 25th, 2020
Journal Published: Frontiers in Pharmacology