Clustering Protein Binding Pockets and Identifying Potential Drug Interactions: A Novel Ligand-Based Featurization Method.

Protein-ligand interactions are essential to drug discovery and drug development efforts. Desirable on-target or multitarget interactions are the first step in finding an effective therapeutic, while undesirable off-target interactions are the first step in assessing safety. In this work, we introduce a novel ligand-based featurization and mapping of human protein pockets to identify closely related protein targets and to project novel drugs into a hybrid protein-ligand feature space to identify their likely protein interactions. Using structure-based template matches from PDB, protein pockets are featured by the ligands that bind to their best co-complex template matches. The simplicity and interpretability of this approach provide a granular characterization of the human proteome at the protein-pocket level instead of the traditional protein-level characterization by family, function, or pathway. We demonstrate the power of this featurization method by clustering a subset of the human proteome and evaluating the predicted cluster associations of over 7000 compounds.

OFraMP: a fragment-based tool to facilitate the parametrization of large molecules.

An Online tool for Fragment-based Molecule Parametrization (OFraMP) is described. OFraMP is a web application for assigning atomic interaction parameters to large molecules by matching sub-fragments within the target molecule to equivalent sub-fragments within the Automated Topology Builder (ATB, atb.uq.edu.au) database. OFraMP identifies and compares alternative molecular fragments from the ATB database, which contains over 890,000 pre-parameterized molecules, using a novel hierarchical matching procedure. Atoms are considered within the context of an extended local environment (buffer region) with the degree of similarity between an atom in the target molecule and that in the proposed match controlled by varying the size of the buffer region. Adjacent matching atoms are combined into progressively larger matched sub-structures. The user then selects the most appropriate match. OFraMP also allows users to manually alter interaction parameters and automates the submission of missing substructures to the ATB in order to generate parameters for atoms in environments not represented in the existing database. The utility of OFraMP is illustrated using the anti-cancer agent paclitaxel and a dendrimer used in organic semiconductor devices. OFraMP applied to paclitaxel (ATB ID 35922).

Machine Learning Models to Predict Inhibition of the Bile Salt Export Pump.

Cholestatic liver injury is frequently associated with drug inhibition of bile salt transporters, such as the bile salt export pump (BSEP). Reliable in silico models to predict BSEP inhibition directly from chemical structures would significantly reduce costs during drug discovery and could help avoid injury to patients. We report our development of classification and regression models for BSEP inhibition with substantially improved performance over previously published models. We assessed the performance effects of different methods of chemical featurization, data set partitioning, and class labeling and identified the methods producing models that generalized best to novel chemical entities.

Oximes: inhibitors of human recombinant acetylcholinesterase. A structure-activity relationship (SAR) study.

Acetylcholinesterase (AChE) reactivators were developed for the treatment of organophosphate intoxication. Standard care involves the use of anticonvulsants (e.g., diazepam), parasympatolytics (e.g., atropine) and oximes that restore AChE activity. However, oximes also bind to the active site of AChE, simultaneously acting as reversible inhibitors. The goal of the present study is to determine how oxime structure influences the inhibition of human recombinant AChE (hrAChE). Therefore, 24 structurally different oximes were tested and the results compared to the previous eel AChE (EeAChE) experiments. Structural factors that were tested included the number of pyridinium rings, the length and structural features of the linker, and the number and position of the oxime group on the pyridinium ring.

Advances in Computational Approaches for Estimating Passive Permeability in Drug Discovery.

Passive permeation of cellular membranes is a key feature of many therapeutics. The relevance of passive permeability spans all biological systems as they all employ biomembranes for compartmentalization. A variety of computational techniques are currently utilized and under active development to facilitate the characterization of passive permeability. These methods include lipophilicity relations, molecular dynamics simulations, and machine learning, which vary in accuracy, complexity, and computational cost. This review briefly introduces the underlying theories, such as the prominent inhomogeneous solubility diffusion model, and covers a number of recent applications. Various machine-learning applications, which have demonstrated good potential for high-volume, data-driven permeability predictions, are also discussed. Due to the confluence of novel computational methods and next-generation exascale computers, we anticipate an exciting future for computationally driven permeability predictions.

Evaluating point-prediction uncertainties in neural networks for protein-ligand binding prediction.

Neural Network (NN) models provide potential to speed up the drug discovery process and reduce its failure rates. The success of NN models requires uncertainty quantification (UQ) as drug discovery explores chemical space beyond the training data distribution. Standard NN models do not provide uncertainty information. Some methods require changing the NN architecture or training procedure, limiting the selection of NN models. Moreover, predictive uncertainty can come from different sources. It is important to have the ability to separately model different types of predictive uncertainty, as the model can take assorted actions depending on the source of uncertainty. In this paper, we examine UQ methods that estimate different sources of predictive uncertainty for NN models aiming at protein-ligand binding prediction. We use our prior knowledge on chemical compounds to design the experiments. By utilizing a visualization method we create non-overlapping and chemically diverse partitions from a collection of chemical compounds. These partitions are used as training and test set splits to explore NN model uncertainty. We demonstrate how the uncertainties estimated by the selected methods describe different sources of uncertainty under different partitions and featurization schemes and the relationship to prediction error.

Development of a CNS-permeable reactivator for nerve agent exposure: an iterative, multi-disciplinary approach.

Nerve agents have experienced a resurgence in recent times with their use against civilian targets during the attacks in Syria (2012), the poisoning of Sergei and Yulia Skripal in the United Kingdom (2018) and Alexei Navalny in Russia (2020), strongly renewing the importance of antidote development against these lethal substances. The current standard treatment against their effects relies on the use of small molecule-based oximes that can efficiently restore acetylcholinesterase (AChE) activity. Despite their efficacy in reactivating AChE, the action of drugs like 2-pralidoxime (2-PAM) is primarily limited to the peripheral nervous system (PNS) and, thus, provides no significant protection to the central nervous system (CNS). This lack of action in the CNS stems from their ionic nature that, on one end makes them very powerful reactivators and on the other renders them ineffective at crossing the Blood Brain Barrier (BBB) to reach the CNS. In this report, we describe the use of an iterative approach composed of parallel chemical and in silico syntheses, computational modeling, and a battery of detailed in vitro and in vivo assays that resulted in the identification of a promising, novel CNS-permeable oxime reactivator. Additional experiments to determine acute and chronic toxicity are ongoing.

Discovery of Small-Molecule Inhibitors of SARS-CoV-2 Proteins Using a Computational and Experimental Pipeline.

A rapid response is necessary to contain emergent biological outbreaks before they can become pandemics. The novel coronavirus (SARS-CoV-2) that causes COVID-19 was first reported in December of 2019 in Wuhan, China and reached most corners of the globe in less than two months. In just over a year since the initial infections, COVID-19 infected almost 100 million people worldwide. Although similar to SARS-CoV and MERS-CoV, SARS-CoV-2 has resisted treatments that are effective against other coronaviruses. Crystal structures of two SARS-CoV-2 proteins, spike protein and main protease, have been reported and can serve as targets for studies in neutralizing this threat. We have employed molecular docking, molecular dynamics simulations, and machine learning to identify from a library of 26 million molecules possible candidate compounds that may attenuate or neutralize the effects of this virus. The viability of selected candidate compounds against SARS-CoV-2 was determined experimentally by biolayer interferometry and FRET-based activity protein assays along with virus-based assays. In the pseudovirus assay, imatinib and lapatinib had IC50 values below 10 µM, while candesartan cilexetil had an IC50 value of approximately 67 µM against Mpro in a FRET-based activity assay. Comparatively, candesartan cilexetil had the highest selectivity index of all compounds tested as its half-maximal cytotoxicity concentration 50 (CC50) value was the only one greater than the limit of the assay (>100 µM).

Simulations of macromolecules in protective and denaturing osmolytes: properties of mixed solvent systems and their effects on water and protein structure and dynamics.

Rarely is any solution simply solute and water. In vivo, solutes, such as proteins and nucleic acids, swim in a sea of water, salts, ions, small molecules, and lipids, not to mention other macromolecules. In vitro, virtually all solutions contain a mixture of aqueous solvents, or "cosolvents" [i.e., solvent(s) in addition to water], that can alter the dynamics, behavior, solubility, and stability of proteins and nucleic acids. We have developed models for a number of cosolvents, including the denaturant urea and the small chemical chaperone trimethylamine N-oxide (TMAO). This chapter examines the models for these two cosolvents in the context of experimental data. The direct and indirect effects of these molecules on water and protein are studied with molecular dynamics simulations. These observations and conclusions are drawn from simulations of these molecules in pure water and as a cosolvent for the protein chymotrypsin inhibitor 2. Urea-induced denaturation occurs initially through attack of the protein by water and hydration of hydrophobic protein moieties as a result of disruption of the hydrogen bonding network of water by urea. This indirect denaturing effect of urea is followed by more direct action as urea replaces some waters involved in the initial hydration of the hydrophobic core and subsequently binds to polar residues and the protein main chain to compete with the intraprotein hydrogen bonds. In the case of TMAO, we find that it encourages water-water interactions, thereby stabilizing the protein as a result of the increased penalty for the hydration of hydrophobic residues.

The biodistribution and pharmacokinetics of the oxime acetylcholinesterase reactivator RS194B in guinea pigs.

Organophosphorus-based (OP) nerve agents represent some of the most toxic substances known to mankind. The current standard of care for exposure has changed very little in the past decades, and relies on a combination of atropine to block receptor activity and oxime-type acetylcholinesterase (AChE) reactivators to reverse the OP binding to AChE. Although these oximes can block the effects of nerve agents, their overall efficacy is reduced by their limited capacity to cross the blood-brain barrier (BBB). RS194B, a new oxime developed by Radic et al. (J. Biol. Chem., 2012) has shown promise for enhanced ability to cross the BBB. To fully assess the potential of this compound as an effective treatment for nerve agent poisoning, a comprehensive evaluation of its pharmacokinetic (PK) and biodistribution profiles was performed using both intravenous and intramuscular exposure routes. The ultra-sensitive technique of accelerator mass spectrometry was used to quantify the compound's PK profile, tissue distribution, and brain/plasma ratio at four dose concentrations in guinea pigs. PK analysis revealed a rapid distribution of the oxime with a plasma t1/2 of ∼1 h. Kidney and liver had the highest concentrations per gram of tissue followed by lung, spleen, heart and brain for all dose concentrations tested. The Cmax in the brain ranged between 0.03 and 0.18% of the administered dose, and the brain-to-plasma ratio ranged from 0.04 at the 10 mg/kg dose to 0.18 at the 200 mg/kg dose demonstrating dose dependent differences in brain and plasma concentrations. In vitro studies show that both passive diffusion and active transport contribute little to RS194B traversal of the BBB. These results indicate that biodistribution is widespread, but very low quantities accumulate in the guinea pig brain, indicating this compound may not be suitable as a centrally active reactivator.

Predicting a Drug's Membrane Permeability: A Computational Model Validated With in Vitro Permeability Assay Data.

Membrane permeability is a key property to consider during the drug design process, and particularly vital when dealing with small molecules that have intracellular targets as their efficacy highly depends on their ability to cross the membrane. In this work, we describe the use of umbrella sampling molecular dynamics (MD) computational modeling to comprehensively assess the passive permeability profile of a range of compounds through a lipid bilayer. The model was initially calibrated through in vitro validation studies employing a parallel artificial membrane permeability assay (PAMPA). The model was subsequently evaluated for its quantitative prediction of permeability profiles for a series of custom synthesized and closely related compounds. The results exhibited substantially improved agreement with the PAMPA data, relative to alternative existing methods. Our work introduces a computational model that underwent progressive molding and fine-tuning as a result of its synergistic collaboration with numerous in vitro PAMPA permeability assays. The presented computational model introduces itself as a useful, predictive tool for permeability prediction.

Computational Enzymology and Organophosphorus Degrading Enzymes: Promising Approaches Toward Remediation Technologies of Warfare Agents and Pesticides.

The re-emergence of chemical weapons as a global threat in hands of terrorist groups, together with an increasing number of pesticides intoxications and environmental contaminations worldwide, has called the attention of the scientific community for the need of improvement in the technologies for detoxification of organophosphorus (OP) compounds. A compelling strategy is the use of bioremediation by enzymes that are able to hydrolyze these molecules to harmless chemical species. Several enzymes have been studied and engineered for this purpose. However, their mechanisms of action are not well understood. Theoretical investigations may help elucidate important aspects of these mechanisms and help in the development of more efficient bio-remediators. In this review, we point out the major contributions of computational methodologies applied to enzyme based detoxification of OPs. Furthermore, we highlight the use of PTE, PON, DFP, and BuChE as enzymes used in OP detoxification process and how computational tools such as molecular docking, molecular dynamics simulations and combined quantum mechanical/molecular mechanics have and will continue to contribute to this very important area of research.

Increasing temperature accelerates protein unfolding without changing the pathway of unfolding.

We have traditionally relied on extremely elevated temperatures (498K, 225 degrees C) to investigate the unfolding process of proteins within the timescale available to molecular dynamics simulations with explicit solvent. However, recent advances in computer hardware have allowed us to extend our thermal denaturation studies to much lower temperatures. Here we describe the results of simulations of chymotrypsin inhibitor 2 at seven temperatures, ranging from 298K to 498K. The simulation lengths vary from 94ns to 20ns, for a total simulation time of 344ns, or 0.34 micros. At 298K, the protein is very stable over the full 50ns simulation. At 348K, corresponding to the experimentally observed melting temperature of CI2, the protein unfolds over the first 25ns, explores partially unfolded conformations for 20ns, and then refolds over the last 35ns. Above its melting temperature, complete thermal denaturation occurs in an activated process. Early unfolding is characterized by sliding or breathing motions in the protein core, leading to an unfolding transition state with a weakened core and some loss of secondary structure. After the unfolding transition, the core contacts are rapidly lost as the protein passes on to the fully denatured ensemble. While the overall character and order of events in the unfolding process are well conserved across temperatures, there are substantial differences in the timescales over which these events take place. We conclude that 498K simulations are suitable for elucidating the details of protein unfolding at a minimum of computational expense.

Reproducible radiation-damage processes in proteins irradiated by intense x-ray pulses.

X-ray free-electron lasers have enabled femtosecond protein nanocrystallography, a novel method to determine the structure of proteins. It allows time-resolved imaging of nanocrystals that are too small for conventional crystallography. The short pulse duration helps in overcoming the detrimental effects of radiation damage because x rays are scattered before the sample has been significantly altered. It has been suggested that, fortuitously, the diffraction process self-terminates abruptly once radiation damage destroys the crystalline order. Our calculations show that high-intensity x-ray pulses indeed trigger a cascade of damage processes in ferredoxin crystals, a particular metalloprotein of interest. However, we found that the damage process is initially not completely random. Correlations exist among the protein monomers, so that Bragg diffraction still occurs in the damaged crystals, despite significant atomic displacements. Our results show that the damage process is reproducible to a certain degree, which is potentially beneficial for the orientation step in single-molecule imaging.

Large-Scale First-Principles Molecular Dynamics Simulations with Electrostatic Embedding: Application to Acetylcholinesterase Catalysis.

Enzymes are complicated solvated systems that typically require many atoms to simulate their function with any degree of accuracy. We have recently developed numerical techniques for large scale first-principles molecular dynamics simulations and applied them to the study of the enzymatic reaction catalyzed by acetylcholinesterase. We carried out density functional theory calculations for a quantum-mechanical (QM) subsystem consisting of 612 atoms with an O(N) complexity finite-difference approach. The QM subsystem is embedded inside an external potential field representing the electrostatic effect due to the environment. We obtained finite-temperature sampling by first-principles molecular dynamics for the acylation reaction of acetylcholine catalyzed by acetylcholinesterase. Our calculations show two energy barriers along the reaction coordinate for the enzyme-catalyzed acylation of acetylcholine. The second barrier (8.5 kcal/mol) is rate-limiting for the acylation reaction and in good agreement with experiment.

A wrench in the works of human acetylcholinesterase: soman induced conformational changes revealed by molecular dynamics simulations.

Irreversible inactivation of human acetylcholinesterase (hAChE) by organophosphorous pesticides (OPs) and chemical weapon agents (CWA) has severe morbidity and mortality consequences. We present data from quantum mechanics/molecular mechanics (QM/MM) and 80 classical molecular dynamics (MD) simulations of the apo and soman-adducted forms of hAChE to investigate the effects on the dynamics and protein structure when the catalytic Serine 203 is phosphonylated. We find that the soman phosphonylation of the active site Ser203 follows a water assisted addition-elimination mechanism with the elimination of the fluoride ion being the highest energy barrier at 6.5 kcal/mole. We observe soman-dependent changes in backbone and sidechain motions compared to the apo form of the protein. These alterations restrict the soman-adducted hAChE to a structural state that is primed for the soman adduct to be cleaved and removed from the active site. The altered motions and resulting structures provide alternative pathways into and out of the hAChE active site. In the soman-adducted protein both side and back door pathways are viable for soman adduct access. Correlation analysis of the apo and soman adducted MD trajectories shows that the correlation of gorge entrance and back door motion is disrupted when hAChE is adducted. This supports the hypothesis that substrate and product can use two different pathways as entry and exit sites in the apo form of the protein. These alternative pathways have important implications for the rational design of medical countermeasures.

Adverse drug reaction prediction using scores produced by large-scale drug-protein target docking on high-performance computing machines.

Late-stage or post-market identification of adverse drug reactions (ADRs) is a significant public health issue and a source of major economic liability for drug development. Thus, reliable in silico screening of drug candidates for possible ADRs would be advantageous. In this work, we introduce a computational approach that predicts ADRs by combining the results of molecular docking and leverages known ADR information from DrugBank and SIDER. We employed a recently parallelized version of AutoDock Vina (VinaLC) to dock 906 small molecule drugs to a virtual panel of 409 DrugBank protein targets. L1-regularized logistic regression models were trained on the resulting docking scores of a 560 compound subset from the initial 906 compounds to predict 85 side effects, grouped into 10 ADR phenotype groups. Only 21% (87 out of 409) of the drug-protein binding features involve known targets of the drug subset, providing a significant probe of off-target effects. As a control, associations of this drug subset with the 555 annotated targets of these compounds, as reported in DrugBank, were used as features to train a separate group of models. The Vina off-target models and the DrugBank on-target models yielded comparable median area-under-the-receiver-operating-characteristic-curves (AUCs) during 10-fold cross-validation (0.60-0.69 and 0.61-0.74, respectively). Evidence was found in the PubMed literature to support several putative ADR-protein associations identified by our analysis. Among them, several associations between neoplasm-related ADRs and known tumor suppressor and tumor invasiveness marker proteins were found. A dual role for interstitial collagenase in both neoplasms and aneurysm formation was also identified. These associations all involve off-target proteins and could not have been found using available drug/on-target interaction data. This study illustrates a path forward to comprehensive ADR virtual screening that can potentially scale with increasing number of CPUs to tens of thousands of protein targets and millions of potential drug candidates.

The 108M polymorph of human catechol O-methyltransferase is prone to deformation at physiological temperatures.

The human gene for catechol O-methyltransferase (COMT) contains a common polymorphism that results in substitution of methionine (M) for valine (V) at residue 108 of the soluble form of the protein. While the two proteins have similar kinetic properties, 108M COMT loses activity more rapidly than 108V COMT at 37 degrees C. The cosubstrate S-adenosylmethionine (SAM) stabilizes the activity of 108M COMT at 40 degrees C. The 108M allele has been associated with increased risk for breast cancer, obsessive-compulsive disorder, and aggressive and highly antisocial manifestations of schizophrenia. In the current work, we have constructed homology models for both human COMT polymorphs and performed molecular dynamics simulations of these models at 25, 37, and 50 degrees C to explore the structural consequences of the 108V/M polymorphism. The simulations indicated that replacing valine with the larger methionine residue led to greater solvent exposure of residue 108 and heightened packing interactions between M108 and helices alpha2, alpha4 (especially with R78), and alpha5. These altered packing interactions propagated subtle changes between the polymorphic site and the active site 16 A away, leading to a loosening of the active site. At physiological temperature, 108M COMT sampled a larger distribution of conformations than 108V. 108M COMT was more prone to active-site distortion and had greater overall, and SAM binding site, solvent accessibility than 108V COMT at 37 degrees C. Similar structural perturbations were observed in the 108V protein only at 50 degrees C. Addition of SAM tightened up the cosubstrate pocket in both proteins and prevented the altered packing at the polymorphic site in 108M COMT.

PhIP carcinogenicity in breast cancer: computational and experimental evidence for competitive interactions with human estrogen receptor.

Many carcinogens have been shown to cause tissue specific tumors in animal models. The mechanism for this specificity has not been fully elucidated and is usually attributed to differences in organ metabolism. For heterocyclic amines, potent carcinogens that are formed in well-done meat, the ability to either bind to the estrogen receptor and activate or inhibit an estrogenic response will have a major impact on carcinogenicity. Here, we describe our work with the human estrogen receptor alpha (ERalpha), the mutagenic/carcinogenic heterocyclic amines PhIP, MeIQx, and IFP, and the hydroxylated metabolite of PhIP, N2-hydroxy-PhIP. We demonstrate both by computational docking and NMR analysis that PhIP binds with the ligand binding domain (LBD). This binding competes with estradiol (E2) in the native E2 binding cavity of the receptor. In vitro assays show that PhIP, in contrast to the other heterocyclic amines, increases cell proliferation in MCF-7 human breast cancer cells and activates the ERalpha receptor. We also find that other heterocyclic amines and N2-hydroxy-PhIP inhibit ERalpha activation. We propose that the mechanism for the tissue-specific carcinogenicity seen in the rat breast tumors and the presumptive human breast cancer associated with the consumption of well-done meat maybe mediated by this receptor activation.

Counteraction of urea-induced protein denaturation by trimethylamine N-oxide: a chemical chaperone at atomic resolution.

Proteins are very sensitive to their solvent environments. Urea is a common chemical denaturant of proteins, yet some animals contain high concentrations of urea. These animals have evolved an interesting mechanism to counteract the effects of urea by using trimethylamine N-oxide (TMAO). The molecular basis for the ability of TMAO to act as a chemical chaperone remains unknown. Here, we describe molecular dynamics simulations of a small globular protein, chymotrypsin inhibitor 2, in 8 M urea and 4 M TMAO/8 M urea solutions, in addition to other control simulations, to investigate this effect at the atomic level. In 8 M urea, the protein unfolds, and urea acts in both a direct and indirect manner to achieve this effect. In contrast, introduction of 4 M TMAO counteracts the effect of urea and the protein remains well structured. TMAO makes few direct interactions with the protein. Instead, it prevents unfolding of the protein by structuring the solvent. In particular, TMAO orders the solvent and discourages it from competing with intraprotein H bonds and breaking up the hydrophobic core of the protein.