Montage electron tomography of vitrified specimens.

Cryo-electron tomography provides detailed views of macromolecules in situ. However, imaging a large field of view to provide more cellular context requires reducing magnification during data collection, which in turn restricts the resolution. To circumvent this trade-off between field of view and resolution, we have developed a montage data collection scheme that uniformly distributes the dose throughout the specimen. In this approach, sets of slightly overlapping circular tiles are collected at high magnification and stitched to form a composite projection image at each tilt angle. These montage tilt-series are then reconstructed into massive tomograms with a small pixel size but a large field of view. For proof-of-principle, we applied this method to the thin edge of HeLa cells. Thon rings to better than 10 Å were detected in the montaged tilt-series, and diverse cellular features were observed in the resulting tomograms. These results indicate that the additional dose required by this technique is not prohibitive to performing structural analysis to intermediate resolution across a large field of view. We anticipate that montage tomography will prove particularly useful for lamellae, increase the likelihood of imaging rare cellular events, and facilitate visual proteomics.

Exploring Deep Learning of Quantum Chemical Properties for Absorption, Distribution, Metabolism, and Excretion Predictions.

Quantum mechanical (QM) descriptors of small molecules have wide applicability in understanding organic reactivity and molecular properties, but the substantial compute cost required for ab initio QM calculations limits their broad usage. Here, we investigate the use of deep learning for predicting QM descriptors, with the goal of enabling usage of near-QM accuracy electronic properties on large molecular data sets such as those seen in drug discovery. Several deep learning approaches have previously been benchmarked on a published data set called QM9, where 12 ground-state properties have been calculated for molecules with up to nine heavy atoms, limited to C, H, N, O, and F elements. To advance the work beyond the QM9 chemical space and enable application to molecules encountered in drug discovery, we extend the QM9 data set by creating a QM9-extended data set covering an additional ∼20,000 molecules containing S and Cl atoms. Using this extended set, we generate new deep learning models as well as leverage ANI-2x models to provide predictions on larger, more diverse molecules common in drug discovery, and we find the models estimate 11 of 12 ground-state properties reasonably. We use the predicted QM descriptors to augment graph convolutional neural network (GCNN) models for selected ADME end points (rat microsomal clearance, hepatic clearance, total clearance, and P-glycoprotein efflux) and found varying degrees of performance improvement compared to nonaugmented GCNN models, including pronounced improvement in P-glycoprotein efflux prediction.

Exploring PROTAC Cooperativity with Coarse-Grained Alchemical Methods.

Proteolysis targeting chimera (PROTAC) is a novel drug modality that facilitates the degradation of a target protein by inducing proximity with an E3 ligase. In this work, we present a new computational framework to model the cooperativity between PROTAC-E3 binding and PROTAC-target binding principally through protein-protein interactions (PPIs) induced by the PROTAC. Due to the scarcity and low resolution of experimental measurements, the physical and chemical drivers of these non-native PPIs remain to be elucidated. We develop a coarse-grained (CG) approach to model interactions in the target-PROTAC-E3 complexes, which enables converged thermodynamic estimations using alchemical free energy calculation methods despite an unconventional scale of perturbations. With minimal parametrization, we successfully capture fundamental principles of cooperativity, including the optimality of intermediate PROTAC linker lengths that originates from configurational entropy. We qualitatively characterize the dependency of cooperativity on PROTAC linker lengths and protein charges and shapes. Minimal inclusion of sequence- and conformation-specific features in our current force field, however, limits quantitative modeling to reproduce experimental measurements, but further development of the CG model may allow for efficient computational screening to optimize PROTAC cooperativity.

PotentialNet for Molecular Property Prediction.

The arc of drug discovery entails a multiparameter optimization problem spanning vast length scales. The key parameters range from solubility (angstroms) to protein-ligand binding (nanometers) to in vivo toxicity (meters). Through feature learning-instead of feature engineering-deep neural networks promise to outperform both traditional physics-based and knowledge-based machine learning models for predicting molecular properties pertinent to drug discovery. To this end, we present the PotentialNet family of graph convolutions. These models are specifically designed for and achieve state-of-the-art performance for protein-ligand binding affinity. We further validate these deep neural networks by setting new standards of performance in several ligand-based tasks. In parallel, we introduce a new metric, the Regression Enrichment Factor EFχ (R), to measure the early enrichment of computational models for chemical data. Finally, we introduce a cross-validation strategy based on structural homology clustering that can more accurately measure model generalizability, which crucially distinguishes the aims of machine learning for drug discovery from standard machine learning tasks.