• MoleculeNet: A Benchmark for Molecular Machine Learning(1703.00564)

Oct. 26, 2018 physics.chem-ph, cs.LG, stat.ML
Molecular machine learning has been maturing rapidly over the last few years. Improved methods and the presence of larger datasets have enabled machine learning algorithms to make increasingly accurate predictions about molecular properties. However, algorithmic progress has been limited due to the lack of a standard benchmark to compare the efficacy of proposed methods; most new algorithms are benchmarked on different datasets making it challenging to gauge the quality of proposed methods. This work introduces MoleculeNet, a large scale benchmark for molecular machine learning. MoleculeNet curates multiple public datasets, establishes metrics for evaluation, and offers high quality open-source implementations of multiple previously proposed molecular featurization and learning algorithms (released as part of the DeepChem open source library). MoleculeNet benchmarks demonstrate that learnable representations are powerful tools for molecular machine learning and broadly offer the best performance. However, this result comes with caveats. Learnable representations still struggle to deal with complex tasks under data scarcity and highly imbalanced classification. For quantum mechanical and biophysical datasets, the use of physics-aware featurizations can be more important than choice of particular learning algorithm.
• Binding Pathway of Opiates to $\mu$ Opioid Receptors Revealed by Unsupervised Machine Learning(1804.08206)

April 23, 2018 q-bio.QM, q-bio.BM
Many important analgesics relieve pain by binding to the $\mu$-Opioid Receptor ($\mu$OR), which makes the $\mu$OR among the most clinically relevant proteins of the G Protein Coupled Receptor (GPCR) family. Despite previous studies on the activation pathways of the GPCRs, the mechanism of opiate binding and the selectivity of $\mu$OR are largely unknown. We performed extensive molecular dynamics (MD) simulation and analysis to find the selective allosteric binding sites of the $\mu$OR and the path opiates take to bind to the orthosteric site. In this study, we predicted that the allosteric site is responsible for the attraction and selection of opiates. Using Markov state models and machine learning, we traced the pathway of opiates in binding to the orthosteric site, the main binding pocket. Our results have important implications in designing novel analgesics.
• Machine Learning Harnesses Molecular Dynamics to Discover New $\mu$ Opioid Chemotypes(1803.04479)

March 12, 2018 q-bio.BM, stat.ML
Computational chemists typically assay drug candidates by virtually screening compounds against crystal structures of a protein despite the fact that some targets, like the $\mu$ Opioid Receptor and other members of the GPCR family, traverse many non-crystallographic states. We discover new conformational states of $\mu OR$ with molecular dynamics simulation and then machine learn ligand-structure relationships to predict opioid ligand function. These artificial intelligence models identified a novel $\mu$ opioid chemotype.
• Spatial Graph Convolutions for Drug Discovery(1803.04465)

March 12, 2018 cs.LG
Predicting the binding free energy, or affinity, of a small molecule for a protein target is frequently the first step along the arc of drug discovery. High throughput experimental and virtual screening both suffer from low accuracy, whereas more accurate approaches in both domains suffer from lack of scale due to either financial or temporal constraints. While machine learning (ML) has made immense progress in the fields of computer vision and natural language processing, it has yet to offer comparable improvements over domain-expertise driven algorithms in the molecular sciences. In this paper, we propose new Deep Neural Network (DNN) architectures for affinity prediction. The new model architectures are at least competitive with, and in many cases state-of-the-art compared to previous knowledge-based and physics-based approaches. In addition to more standard evaluation metrics, we also propose the Regression Enrichment Factor $EF_\chi^{(R)}$ for the community to benchmark against in future affinity prediction studies. Finally, we suggest the adaptation of an agglomerative clustering cross-validation strategy to more accurately reflect the generalization capacity of ML-based affinity models in future works.
• Atomic Convolutional Networks for Predicting Protein-Ligand Binding Affinity(1703.10603)

March 30, 2017 physics.chem-ph, cs.LG, stat.ML
Empirical scoring functions based on either molecular force fields or cheminformatics descriptors are widely used, in conjunction with molecular docking, during the early stages of drug discovery to predict potency and binding affinity of a drug-like molecule to a given target. These models require expert-level knowledge of physical chemistry and biology to be encoded as hand-tuned parameters or features rather than allowing the underlying model to select features in a data-driven procedure. Here, we develop a general 3-dimensional spatial convolution operation for learning atomic-level chemical interactions directly from atomic coordinates and demonstrate its application to structure-based bioactivity prediction. The atomic convolutional neural network is trained to predict the experimentally determined binding affinity of a protein-ligand complex by direct calculation of the energy associated with the complex, protein, and ligand given the crystal structure of the binding pose. Non-covalent interactions present in the complex that are absent in the protein-ligand sub-structures are identified and the model learns the interaction strength associated with these features. We test our model by predicting the binding free energy of a subset of protein-ligand complexes found in the PDBBind dataset and compare with state-of-the-art cheminformatics and machine learning-based approaches. We find that all methods achieve experimental accuracy and that atomic convolutional networks either outperform or perform competitively with the cheminformatics based methods. Unlike all previous protein-ligand prediction systems, atomic convolutional networks are end-to-end and fully-differentiable. They represent a new data-driven, physics-based deep learning model paradigm that offers a strong foundation for future improvements in structure-based bioactivity prediction.