Search

Background: Artificial intelligence carries the risk of exacerbating some of our most challenging societal problems, but it also has the potential of mitigating and addressing these problems. The confounding effects of race on machine learning is an ongoing subject of research. This study aims to mitigate the impact of race on data-derived models, using an adversarial variational autoencoder (AVAE). In this study, race is a self-reported feature. Race is often excluded as an input variable, however, due to the high correlation between race and several other variables, race is still implicitly encoded in the data., Methods: We propose building a model that (1) learns a low dimensionality latent spaces, (2) employs an adversarial training procedure that ensure its latent space does not encode race, and (3) contains necessary information for reconstructing the data. We train the autoencoder to ensure the latent space does not indirectly encode race., Results: In this study, AVAE successfully removes information about race from the latent space (AUC ROC = 0.5). In contrast, latent spaces constructed using other approaches still allow the reconstruction of race with high fidelity. The AVAE's latent space does not encode race but conveys important information required to reconstruct the dataset. Furthermore, the AVAE's latent space does not predict variables related to race (R 2 = 0.003), while a model that includes race does (R 2 = 0.08)., Conclusions: Though we constructed a race-independent latent space, any variable could be similarly controlled. We expect AVAEs are one of many approaches that will be required to effectively manage and understand bias in ML., (© 2024. The Author(s).)

Big genomic data and artificial intelligence (AI) are ushering in an era of precision medicine, providing opportunities to study previously under-represented subtypes and rare diseases rather than categorize them as variances. However, clinical researchers face challenges in accessing such novel technologies as well as reliable methods to study small datasets or subcohorts with unique phenotypes. To address this need, we developed an integrative approach, GAiN, to capture patterns of gene expression from small datasets on the basis of an ensemble of generative adversarial networks (GANs) while leveraging big population data. Where conventional biostatistical methods fail, GAiN reliably discovers differentially expressed genes (DEGs) and enriched pathways between two cohorts with limited numbers of samples (n = 10) when benchmarked against a gold standard. GAiN is freely available at GitHub. Thus, GAiN may serve as a crucial tool for gene expression analysis in scenarios with limited samples, as in the context of rare diseases, under-represented populations, or limited investigator resources., Competing Interests: The authors declare no competing interests., (© 2023 The Author(s).)

Background: Race is commonly used as a proxy for multiple features including socioeconomic status. It is critical to dissociate these factors, to identify mechanisms that affect infant outcomes, such as birth weight, gestational age, and brain development, and to direct appropriate interventions and shape public policy., Methods: Demographic, socioeconomic, and clinical variables were used to model infant outcomes. There were 351 participants included in the analysis for birth weight and gestational age. For the analysis using brain volumes, 280 participants were included after removing participants with missing magnetic resonance imaging scans and those matching our exclusion criteria. We modeled these three different infant outcomes, including infant brain, birth weight, and gestational age, with both linear and nonlinear models., Results: Nonlinear models were better predictors of infant birth weight than linear models ( R 2  = 0.172 vs. R 2  = 0.145, p  = .005). In contrast to linear models, nonlinear models ranked income, neighborhood disadvantage, and experiences of discrimination higher in importance than race while modeling birth weight. Race was not an important predictor for either gestational age or structural brain volumes., Conclusions: Consistent with the extant social science literature, the findings related to birth weight suggest that race is a linear proxy for nonlinear factors related to structural racism. Methods that can disentangle factors often correlated with race are important for policy in that they may better identify and rank the modifiable factors that influence outcomes., (© 2023 The Authors.)

Cytochrome P450 enzymes aid in the elimination of a preponderance of small molecule drugs, but can generate reactive metabolites that may adversely react with protein and DNA and prompt drug candidate attrition or market withdrawal. Previously developed models help understand how these enzymes modify molecule structure by predicting sites of metabolism or characterizing formation of metabolite-biomolecule adducts. However, the majority of reactive metabolites are formed by multiple metabolic steps, and understanding the progenitor molecule's network-level behavior necessitates an integrative approach that blends multiple site of metabolism and structure inference models. Our previously developed tool, XenoNet 1.0, generates metabolic networks, where nodes are molecules and weighted edges are metabolic transformations. We extend XenoNet with a bidirectional message passing neural network that integrates edge feature information and local network structure using edge-conditioned graph convolutions and jumping knowledge to predict the authenticity of inferred Phase I metabolite structures. Our model significantly outperformed prior work and algorithmic baselines on a data set of 311 networks and 6606 intermediates annotated using a chemically diverse set of 20 736 individual in vitro and in vivo reaction records accounting for 92.3% of all human Phase I metabolism in the Accelrys Metabolite Database. Cross-validated predictions resulted in area under the receiver operating characteristic curves of 88.5% and 87.6% for separating experimentally observed and unobserved metabolites at global and network levels, respectively. Further analysis verified robustness to networks of varying depth and breadth, accurate detection of metabolites, such as d,l-methamphetamine, that are experimentally observed or unobserved in different network contexts, extraction of important metabolic subnetworks, and identification of known bioactivation pathways, such as for nimesulide and terbinafine. By exploiting network structures, our approach accurately suggests unreported metabolites for experimental study and may rationalize modifications for avoiding deleterious pathways antecedent to reactive metabolite formation.

The gut microbiota in early childhood is linked to asthma risk, but may continue to affect older patients with asthma. Here, we profile the gut microbiota of 38 children (19 asthma, median age 8) and 57 adults (17 asthma, median age 28) by 16S rRNA sequencing and find individuals with asthma harbored compositional differences from healthy controls in both adults and children. We develop a model to aid the design of mechanistic experiments in gnotobiotic mice and show enterotoxigenic Bacteroides fragilis (ETBF) is more prevalent in the gut microbiota of patients with asthma compared to healthy controls. In mice, ETBF, modulated by community context, can increase oxidative stress in the lungs during allergic airway inflammation (AAI). Our results provide evidence that ETBF affects the phenotype of airway inflammation in a subset of patients with asthma which suggests that therapies targeting the gut microbiota may be helpful tools for asthma control., Competing Interests: The authors have no competing interests to declare., (© 2023 The Author(s).)

An accurate estimation of glomerular filtration rate (GFR) is clinically crucial for kidney disease diagnosis and predicting the prognosis of chronic kidney disease (CKD). Machine learning methodologies such as deep neural networks provide a potential avenue for increasing accuracy in GFR estimation. We developed a novel deep learning architecture, a deep and shallow neural network, to estimate GFR (dlGFR for short) and examined its comparative performance with estimated GFR from Modification of Diet in Renal Disease (MDRD) and Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations. The dlGFR model jointly trains a shallow learning model and a deep neural network to enable both linear transformation from input features to a log GFR target, and non-linear feature embedding for stage of kidney function classification. We validate the proposed methods on the data from multiple studies obtained from the NIDDK Central Database Repository. The deep learning model predicted values of GFR within 30% of measured GFR with 88.3% accuracy, compared to the 87.1% and 84.7% of the accuracy achieved by CKD-EPI and MDRD equations (p = 0.051 and p < 0.001, respectively). Our results suggest that deep learning methods are superior to equations resulting from traditional statistical methods in estimating glomerular filtration rate. Based on these results, an end-to-end predication system has been deployed to facilitate use of the proposed dlGFR algorithm.

Purpose of Review: The field of pathology is currently undergoing a significant transformation from traditional glass slides to a digital format dependent on whole slide imaging. Transitioning from glass to digital has opened the field to development and application of image analysis technology, commonly deep learning methods (artificial intelligence [AI]) to assist pathologists with tissue examination. Nephropathology is poised to leverage this technology to improve precision, accuracy, and efficiency in clinical practice., Recent Findings: Through a multidisciplinary approach, nephropathologists, and computer scientists have made significant recent advances in developing AI technology to identify histological structures within whole slide images (segmentation), quantification of histologic structures, prediction of clinical outcomes, and classifying disease. Virtual staining of tissue and automation of electron microscopy imaging are emerging applications with particular significance for nephropathology., Summary: AI applied to image analysis in nephropathology has potential to transform the field by improving diagnostic accuracy and reproducibility, efficiency, and prognostic power. Reimbursement, demonstration of clinical utility, and seamless workflow integration are essential to widespread adoption., (Copyright © 2022 Wolters Kluwer Health, Inc. All rights reserved.)

Coumadin (R/S-warfarin) anticoagulant therapy is highly efficacious in preventing the formation of blood clots; however, significant inter-individual variations in response risks over or under dosing resulting in adverse bleeding events or ineffective therapy, respectively. Levels of pharmacologically active forms of the drug and metabolites depend on a diversity of metabolic pathways. Cytochromes P450 play a major role in oxidizing R- and S-warfarin to 6-, 7-, 8-, 10-, and 4'-hydroxywarfarin, and warfarin alcohols form through a minor metabolic pathway involving reduction at the C11 position. We hypothesized that due to structural similarities with warfarin, hydroxywarfarins undergo reduction, possibly impacting their pharmacological activity and elimination. We modeled reduction reactions and carried out experimental steady-state reactions with human liver cytosol for conversion of rac -6-, 7-, 8-, 4'-hydroxywarfarin and 10-hydroxywarfarin isomers to the corresponding alcohols. The modeling correctly predicted the more efficient reduction of 10-hydroxywarfarin over warfarin but not the order of the remaining hydroxywarfarins. Experimental studies did not indicate any clear trends in the reduction for rac -hydroxywarfarins or 10-hydroxywarfarin into alcohol 1 and 2. The collective findings indicated the location of the hydroxyl group significantly impacted reduction selectivity among the hydroxywarfarins, as well as the specificity for the resulting metabolites. Based on studies with R- and S-7-hydroxywarfarin, we predicted that all hydroxywarfarin reductions are enantioselective toward R substrates and enantiospecific for S alcohol metabolites. CBR1 and to a lesser extent AKR1C3 reductases are responsible for those reactions. Due to the inefficiency of reactions, only reduction of 10-hydroxywarfarin is likely to be important in clearance of the metabolite. This pathway for 10-hydroxywarfarin may have clinical relevance as well given its anticoagulant activity and capacity to inhibit S -warfarin metabolism., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2022 Pouncey, Barnette, Sinnott, Phillips, Flynn, Hendrickson, Swamidass and Miller.)

Computing quantum chemical properties of small molecules and polymers can provide insights valuable into physicists, chemists, and biologists when designing new materials, catalysts, biological probes, and drugs. Deep learning can compute quantum chemical properties accurately in a fraction of time required by commonly used methods such as density functional theory. Most current approaches to deep learning in quantum chemistry begin with geometric information from experimentally derived molecular structures or pre-calculated atom coordinates. These approaches have many useful applications, but they can be costly in time and computational resources. In this study, we demonstrate that accurate quantum chemical computations can be performed without geometric information by operating in the coordinate-free domain using deep learning on graph encodings. Coordinate-free methods rely only on molecular graphs, the connectivity of atoms and bonds, without atom coordinates or bond distances. We also find that the choice of graph-encoding architecture substantially affects the performance of these methods. The structures of these graph-encoding architectures provide an opportunity to probe an important, outstanding question in quantum mechanics: what types of quantum chemical properties can be represented by local variable models? We find that Wave, a local variable model, accurately calculates the quantum chemical properties, while graph convolutional architectures require global variables. Furthermore, local variable Wave models outperform global variable graph convolution models on complex molecules with large, correlated systems.

Drug-drug interactions account for up to 30% of adverse drug reactions. Increasing prevalence of electronic health records (EHRs) offers a unique opportunity to build machine learning algorithms to identify drug-drug interactions that drive adverse events. In this study, we investigated hospitalizations' data to study drug interactions with non-steroidal anti-inflammatory drugs (NSAIDS) that result in drug-induced liver injury (DILI). We propose a logistic regression based machine learning algorithm that unearths several known interactions from an EHR dataset of about 400,000 hospitalization. Our proposed modeling framework is successful in detecting 87.5% of the positive controls, which are defined by drugs known to interact with diclofenac causing an increased risk of DILI, and correctly ranks aggregate risk of DILI for eight commonly prescribed NSAIDs. We found that our modeling framework is particularly successful in inferring associations of drug-drug interactions from relatively small EHR datasets. Furthermore, we have identified a novel and potentially hepatotoxic interaction that might occur during concomitant use of meloxicam and esomeprazole, which are commonly prescribed together to allay NSAID-induced gastrointestinal (GI) bleeding. Empirically, we validate our approach against prior methods for signal detection on EHR datasets, in which our proposed approach outperforms all the compared methods across most metrics, such as area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC)., Competing Interests: The authors have declared that no competing interests exist.

Diphenylamine NSAIDs are highly prescribed therapeutics for chronic pain despite causing symptomatic hepatotoxicity through mitochondrial damage in five percent of patients taking them. Differences in toxicity are attributed to structural modifications to the diphenylamine scaffold rather than its inherent toxicity. We hypothesize that marketed diphenylamine NSAID substituents affect preference and efficiency of bioactivation pathways and clearance. We parsed the FDA DILIrank hepatotoxicity database and modeled marketed drug bioactivation into quinone-species metabolites to identify a family of seven clinically relevant diphenylamine NSAIDs. These drugs fell into two subgroups, i.e., acetic acid and propionic acid diphenylamines, varying in hepatotoxicity risks and modeled bioactivation propensities. We carried out steady-state kinetic studies to assess bioactivation pathways by trapping quinone-species metabolites with dansyl glutathione. Analysis of the glutathione adducts by mass spectrometry characterized structures while dansyl fluorescence provided quantitative yields for their formation. Resulting kinetics identified four possible bioactivation pathways among the drugs, but reaction preference and efficiency depended upon structural modifications to the diphenylamine scaffold. Strikingly, diphenylamine dihalogenation promotes formation of quinone metabolites through four distinct metabolic pathways with high efficiency, whereas those without aromatic halogen atoms were metabolized less efficiently through two or fewer metabolic pathways. Overall metabolism of the drugs was comparable with bioactivation accounting for 4-13% of clearance. Lastly, we calculated daily bioload exposure of quinone-species metabolites based on bioactivation efficiency, bioavailability, and maximal daily dose. The results revealed stratification into the two subgroups; propionic acid diphenylamines had an average four-fold greater daily bioload compared to acetic acid diphenylamines. However, the lack of sufficient study on the hepatotoxicity for all drugs prevents further correlative analyses. These findings provide critical insights on the impact of diphenylamine bioactivation as a precursor to hepatotoxicity and thus, provide a foundation for better risk assessment in drug discovery and development., (Copyright © 2021. Published by Elsevier B.V.)

The 3,5-dimethylisoxazole motif has become a useful and popular acetyl-lysine mimic employed in isoxazole-containing bromodomain and extra-terminal (BET) inhibitors but may introduce the potential for bioactivations into toxic reactive metabolites. As a test, we coupled deep neural models for quinone formation, metabolite structures, and biomolecule reactivity to predict bioactivation pathways for 32 BET inhibitors and validate the bioactivation of select inhibitors experimentally. Based on model predictions, inhibitors were more likely to undergo bioactivation than reported non-bioactivated molecules containing isoxazoles. The model outputs varied with substituents indicating the ability to scale their impact on bioactivation. We selected OXFBD02, OXFBD04, and I-BET151 for more in-depth analysis. OXFBD's bioactivations were evenly split between traditional quinones and novel extended quinone-methides involving the isoxazole yet strongly favored the latter quinones. Subsequent experimental studies confirmed the formation of both types of quinones for OXFBD molecules, yet traditional quinones were the dominant reactive metabolites. Modeled I-BET151 bioactivations led to extended quinone-methides, which were not verified experimentally. The differences in observed and predicted bioactivations reflected the need to improve overall bioactivation scaling. Nevertheless, our coupled modeling approach predicted BET inhibitor bioactivations including novel extended quinone methides, and we experimentally verified those pathways highlighting potential concerns for toxicity in the development of these new drug leads.

Understanding the structural determinants of a protein's biochemical properties, such as activity and stability, is a major challenge in biology and medicine. Comparing computer simulations of protein variants with different biochemical properties is an increasingly powerful means to drive progress. However, success often hinges on dimensionality reduction algorithms for simplifying the complex ensemble of structures each variant adopts. Unfortunately, common algorithms rely on potentially misleading assumptions about what structural features are important, such as emphasizing larger geometric changes over smaller ones. Here we present DiffNets, self-supervised autoencoders that avoid such assumptions, and automatically identify the relevant features, by requiring that the low-dimensional representations they learn are sufficient to predict the biochemical differences between protein variants. For example, DiffNets automatically identify subtle structural signatures that predict the relative stabilities of β-lactamase variants and duty ratios of myosin isoforms. DiffNets should also be applicable to understanding other perturbations, such as ligand binding.

Meloxicam is a thiazole-containing NSAID that was approved for marketing with favorable clinical outcomes despite being structurally similar to the hepatotoxic sudoxicam. Introduction of a single methyl group on the thiazole results in an overall lower toxic risk, yet the group's impact on P450 isozyme bioactivation is unclear. Through analytical methods, we used inhibitor phenotyping and recombinant P450s to identify contributing P450s, and then measured steady-state kinetics for bioactivation of sudoxicam and meloxicam by the recombinant P450s to determine relative efficiencies. Experiments showed that CYP2C8, 2C19, and 3A4 catalyze sudoxicam bioactivation, and CYP1A2 catalyzes meloxicam bioactivation, indicating that the methyl group not only impacts enzyme affinity for the drugs, but also alters which isozymes catalyze the metabolic pathways. Scaling of relative P450 efficiencies based on average liver concentration revealed that CYP2C8 dominates the sudoxicam bioactivation pathway and CYP2C9 dominates meloxicam detoxification. Dominant P450s were applied for an informatics assessment of electronic health records to identify potential correlations between meloxicam drug-drug interactions and drug-induced liver injury. Overall, our findings provide a cautionary tale on assumed impacts of even simple structural modifications on drug bioactivation while also revealing specific targets for clinical investigations of predictive factors that determine meloxicam-induced idiosyncratic liver injury., Competing Interests: Declaration of Competing Interest The authors report no declarations of interest., (Copyright © 2020 Elsevier B.V. All rights reserved.)

Electrophilically reactive drug metabolites are implicated in many adverse drug reactions. In this mechanism-termed bioactivation-metabolic enzymes convert drugs into reactive metabolites that often conjugate to nucleophilic sites within biological macromolecules like proteins. Toxic metabolite-product adducts induce severe immune responses that can cause sometimes fatal disorders, most commonly in the form of liver injury, blood dyscrasia, or the dermatologic conditions toxic epidermal necrolysis and Stevens-Johnson syndrome. This study models four of the most common metabolic transformations that result in bioactivation: quinone formation, epoxidation, thiophene sulfur-oxidation, and nitroaromatic reduction, by synthesizing models of metabolism and reactivity. First, the metabolism models predict the formation probabilities of all possible metabolites among the pathways studied. Second, the exact structures of these metabolites are enumerated. Third, using these structures, the reactivity model predicts the reactivity of each metabolite. Finally, a feedfoward neural network converts the metabolism and reactivity predictions to a bioactivation prediction for each possible metabolite. These bioactivation predictions represent the joint probability that a metabolite forms and that this metabolite subsequently conjugates to protein or glutathione. Among molecules bioactivated by these pathways, we predicted the correct pathway with an AUC accuracy of 89.98%. Furthermore, the model predicts whether molecules will be bioactivated, distinguishing bioactivated and nonbioactivated molecules with 81.06% AUC. We applied this algorithm to withdrawn drugs. The known bioactivation pathways of alclofenac and benzbromarone were identified by the algorithm, and high probability bioactivation pathways not yet confirmed were identified for safrazine, zimelidine, and astemizole. This bioactivation model-the first of its kind that jointly considers both metabolism and reactivity-enables drug candidates to be quickly evaluated for a toxicity risk that often evades detection during preclinical trials. The XenoSite bioactivation model is available at http://swami.wustl.edu/xenosite/p/bioactivation.

Meclofenamate is a nonsteroidal anti-inflammatory drug used in the treatment of mild-to-moderate pain yet poses a rare risk of hepatotoxicity through an unknown mechanism. Nonsteroidal anti-inflammatory drug (NSAID) bioactivation is a common molecular initiating event for hepatotoxicity. Thus, we hypothesized a similar mechanism for meclofenamate and leveraged computational and experimental approaches to identify and characterize its bioactivation. Analyses employing our XenoNet model indicated possible pathways to meclofenamate bioactivation into 19 reactive metabolites subsequently trapped into glutathione adducts. We describe the first reported bioactivation kinetics for meclofenamate and relative importance of those pathways using human liver microsomes. The findings validated only four of the many bioactivation pathways predicted by modeling. For experimental studies, dansyl glutathione was a critical trap for reactive quinone metabolites and provided a way to characterize adduct structures by mass spectrometry and quantitate yields during reactions. Of the four quinone adducts, we were able to characterize structures for three of them. Based on kinetics, the most efficient bioactivation pathway led to the monohydroxy para-quinone-imine followed by the dechloro-ortho-quinone-imine. Two very inefficient pathways led to the dihydroxy ortho-quinone and a likely multiply adducted quinone. When taken together, bioactivation pathways for meclofenamate accounted for approximately 13% of total metabolism. In sum, XenoNet facilitated prediction of reactive metabolite structures, whereas quantitative experimental studies provided a tractable approach to validate actual bioactivation pathways for meclofenamate. Our results provide a foundation for assessing reactive metabolite load more accurately for future comparative studies with other NSAIDs and drugs in general. SIGNIFICANCE STATEMENT: Meclofenamate bioactivation may initiate hepatotoxicity, yet common risk assessment approaches are often cumbersome and inefficient and yield qualitative insights that do not scale relative bioactivation risks. We developed and applied innovative computational modeling and quantitative kinetics to identify and validate meclofenamate bioactivation pathways and relevance as a function of time and concentration. This strategy yielded novel insights on meclofenamate bioactivation and provides a tractable approach to more accurately and efficiently assess other drug bioactivations and correlate risks to toxicological outcomes., (Copyright © 2021 The Author(s).)

Importance: A chronic shortage of donor kidneys is compounded by a high discard rate, and this rate is directly associated with biopsy specimen evaluation, which shows poor reproducibility among pathologists. A deep learning algorithm for measuring percent global glomerulosclerosis (an important predictor of outcome) on images of kidney biopsy specimens could enable pathologists to more reproducibly and accurately quantify percent global glomerulosclerosis, potentially saving organs that would have been discarded., Objective: To compare the performances of pathologists with a deep learning model on quantification of percent global glomerulosclerosis in whole-slide images of donor kidney biopsy specimens, and to determine the potential benefit of a deep learning model on organ discard rates., Design, Setting, and Participants: This prognostic study used whole-slide images acquired from 98 hematoxylin-eosin-stained frozen and 51 permanent donor biopsy specimen sections retrieved from 83 kidneys. Serial annotation by 3 board-certified pathologists served as ground truth for model training and for evaluation. Images of kidney biopsy specimens were obtained from the Washington University database (retrieved between June 2015 and June 2017). Cases were selected randomly from a database of more than 1000 cases to include biopsy specimens representing an equitable distribution within 0% to 5%, 6% to 10%, 11% to 15%, 16% to 20%, and more than 20% global glomerulosclerosis., Main Outcomes and Measures: Correlation coefficient (r) and root-mean-square error (RMSE) with respect to annotations were computed for cross-validated model predictions and on-call pathologists' estimates of percent global glomerulosclerosis when using individual and pooled slide results. Data were analyzed from March 2018 to August 2020., Results: The cross-validated model results of section images retrieved from 83 donor kidneys showed higher correlation with annotations (r = 0.916; 95% CI, 0.886-0.939) than on-call pathologists (r = 0.884; 95% CI, 0.825-0.923) that was enhanced when pooling glomeruli counts from multiple levels (r = 0.933; 95% CI, 0.898-0.956). Model prediction error for single levels (RMSE, 5.631; 95% CI, 4.735-6.517) was 14% lower than on-call pathologists (RMSE, 6.523; 95% CI, 5.191-7.783), improving to 22% with multiple levels (RMSE, 5.094; 95% CI, 3.972-6.301). The model decreased the likelihood of unnecessary organ discard by 37% compared with pathologists., Conclusions and Relevance: The findings of this prognostic study suggest that this deep learning model provided a scalable and robust method to quantify percent global glomerulosclerosis in whole-slide images of donor kidneys. The model performance improved by analyzing multiple levels of a section, surpassing the capacity of pathologists in the time-sensitive setting of examining donor biopsy specimens. The results indicate the potential of a deep learning model to prevent erroneous donor organ discard.

Atom- or bond-level chemical properties of interest in medicinal chemistry, such as drug metabolism and electrophilic reactivity, are important to understand and predict across arbitrary new molecules. Deep learning can be used to map molecular structures to their chemical properties, but the data sets for these tasks are relatively small, which can limit accuracy and generalizability. To overcome this limitation, it would be preferable to model these properties on the basis of the underlying quantum chemical characteristics of small molecules. However, it is difficult to learn higher level chemical properties from lower level quantum calculations. To overcome this challenge, we pretrained deep learning models to compute quantum chemical properties and then reused the intermediate representations constructed by the pretrained network. Transfer learning, in this way, substantially outperformed models based on chemical graphs alone or quantum chemical properties alone. This result was robust, observable in five prediction tasks: identifying sites of epoxidation by metabolic enzymes and identifying sites of covalent reactivity with cyanide, glutathione, DNA and protein. We see that this approach may substantially improve the accuracy of deep learning models for specific chemical structures, such as aromatic systems.

Adverse drug metabolism often severely impacts patient morbidity and mortality. Unfortunately, drug metabolism experimental assays are costly, inefficient, and slow. Instead, computational modeling could rapidly flag potentially toxic molecules across thousands of candidates in the early stages of drug development. Most metabolism models focus on predicting sites of metabolism (SOMs): the specific substrate atoms targeted by metabolic enzymes. However, SOMs are merely a proxy for metabolic structures: knowledge of an SOM does not explicitly provide the actual metabolite structure. Without an explicit metabolite structure, computational systems cannot evaluate the new molecule's properties. For example, the metabolite's reactivity cannot be automatically predicted, a crucial limitation because reactive drug metabolites are a key driver of adverse drug reactions (ADRs). Additionally, further metabolic events cannot be forecast, even though the metabolic path of the majority of substrates includes two or more sequential steps. To overcome the myopia of the SOM paradigm, this study constructs a well-defined system-termed the metabolic forest-for generating exact metabolite structures. We validate the metabolic forest with the substrate and product structures from a large, chemically diverse, literature-derived dataset of 20 736 records. The metabolic forest finds a pathway linking each substrate and product for 79.42% of these records. By performing a breadth-first search of depth two or three, we improve performance to 88.43 and 88.77%, respectively. The metabolic forest includes a specialized algorithm for producing accurate quinone structures, the most common type of reactive metabolite. To our knowledge, this quinone structure algorithm is the first of its kind, as the diverse mechanisms of quinone formation are difficult to systematically reproduce. We validate the metabolic forest on a previously published dataset of 576 quinone reactions, predicting their structures with a depth three performance of 91.84%. The metabolic forest accurately enumerates metabolite structures, enabling promising new directions such as joint metabolism and reactivity modeling.

Background: Pathologist evaluation of donor liver biopsies provides information for accepting or discarding potential donor livers. Due to the urgent nature of the decision process, this is regularly performed using frozen sectioning at the time of biopsy. The percent steatosis in a donor liver biopsy correlates with transplant outcome, however there is significant inter- and intra-observer variability in quantifying steatosis, compounded by frozen section artifact. We hypothesized that a deep learning model could identify and quantify steatosis in donor liver biopsies., Methods: We developed a deep learning convolutional neural network that generates a steatosis probability map from an input whole slide image (WSI) of a hematoxylin and eosin-stained frozen section, and subsequently calculates the percent steatosis. Ninety-six WSI of frozen donor liver sections from our transplant pathology service were annotated for steatosis and used to train (n = 30 WSI) and test (n = 66 WSI) the deep learning model., Findings: The model had good correlation and agreement with the annotation in both the training set (r of 0.88, intraclass correlation coefficient [ICC] of 0.88) and novel input test sets (r = 0.85 and ICC=0.85). These measurements were superior to the estimates of the on-service pathologist at the time of initial evaluation (r = 0.52 and ICC=0.52 for the training set, and r = 0.74 and ICC=0.72 for the test set)., Interpretation: Use of this deep learning algorithm could be incorporated into routine pathology workflows for fast, accurate, and reproducible donor liver evaluation., Funding: Mid-America Transplant Society., (Copyright © 2020 The Authors. Published by Elsevier B.V. All rights reserved.)

Drug metabolism is a common cause of adverse drug reactions. Drug molecules can be metabolized into reactive metabolites, which can conjugate to biomolecules, like protein and DNA, in a process termed bioactivation. To mitigate adverse reactions caused by bioactivation, both experimental and computational screening assays are utilized. Experimental assays for assessing the formation of reactive metabolites are low throughput and expensive to perform, so they are often reserved until later stages of the drug development pipeline when the drug candidate pools are already significantly narrowed. In contrast, computational methods are high throughput and cheap to perform to screen thousands to millions of compounds for potentially toxic molecules during the early stages of the drug development pipeline. Commonly used computational methods focus on detecting and structurally characterizing reactive metabolite-biomolecule adducts or predicting sites on a drug molecule that are liable to form reactive metabolites. However, such methods are often only concerned with the structure of the initial drug molecule or of the adduct formed when a biomolecule conjugates to a reactive metabolite. Thus, these methods are likely to miss intermediate metabolites that may lead to subsequent reactive metabolite formation. To address these shortcomings, we create XenoNet, a metabolic network predictor, that can take a pair of a substrate and a target product as input and (1) enumerate pathways, or sequences of intermediate metabolite structures, between the pair, and (2) compute the likelihood of those pathways and intermediate metabolites. We validate XenoNet on a large, chemically diverse data set of 17 054 metabolic networks built from a literature-derived reaction database. Each metabolic network has a defined substrate molecule that has been experimentally observed to undergo metabolism into a defined product metabolite. XenoNet can predict experimentally observed pathways and intermediate metabolites linking the input substrate and product pair with a recall of 88 and 46%, respectively. Using likelihood scoring, XenoNet also achieves a top-one pathway and intermediate metabolite accuracy of 93.6 and 51.9%, respectively. We further validate XenoNet against prior methods for metabolite prediction. XenoNet significantly outperforms all prior methods across multiple metrics. XenoNet is available at https://swami.wustl.edu/xenonet.

Thiazoles are biologically active aromatic heterocyclic rings occurring frequently in natural products and drugs. These molecules undergo typically harmless elimination; however, a hepatotoxic response can occur due to multistep bioactivation of the thiazole to generate a reactive thioamide. A basis for those differences in outcomes remains unknown. A textbook example is the high hepatotoxicity observed for sudoxicam in contrast to the relative safe use and marketability of meloxicam, which differs in structure from sudoxicam by the addition of a single methyl group. Both drugs undergo bioactivation, but meloxicam exhibits an additional detoxification pathway due to hydroxylation of the methyl group. We hypothesized that thiazole bioactivation efficiency is similar between sudoxicam and meloxicam due to the methyl group being a weak electron donator, and thus, the relevance of bioactivation depends on the competing detoxification pathway. For a rapid analysis, we modeled epoxidation of sudoxicam derivatives to investigate the impact of substituents on thiazole bioactivation. As expected, electron donating groups increased the likelihood for epoxidation with a minimal effect for the methyl group, but model predictions did not extrapolate well among all types of substituents. Through analytical methods, we measured steady-state kinetics for metabolic bioactivation of sudoxicam and meloxicam by human liver microsomes. Sudoxicam bioactivation was 6-fold more efficient than that for meloxicam, yet meloxicam showed a 6-fold higher efficiency of detoxification than bioactivation. Overall, sudoxicam bioactivation was 15-fold more likely than meloxicam considering all metabolic clearance pathways. Kinetic differences likely arise from different enzymes catalyzing respective metabolic pathways based on phenotyping studies. Rather than simply providing an alternative detoxification pathway, the meloxicam methyl group suppressed the bioactivation reaction. These findings indicate the impact of thiazole substituents on bioactivation is more complex than previously thought and likely contributes to the unpredictability of their toxic potential., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2020. Published by Elsevier B.V.)

Metabolism of drugs affects their absorption, distribution, efficacy, excretion, and toxicity profiles. Metabolism is routinely assessed experimentally using recombinant enzymes, human liver microsome, and animal models. Unfortunately, these experiments are expensive, time-consuming, and often extrapolate poorly to humans because they fail to capture the full breadth of metabolic reactions observed in vivo . As a result, metabolic pathways leading to the formation of toxic metabolites are often missed during drug development, giving rise to costly failures. To address some of these limitations, computational metabolism models can rapidly and cost-effectively predict sites of metabolism-the atoms or bonds which undergo enzymatic modifications-on thousands of drug candidates, thereby improving the likelihood of discovering metabolic transformations forming toxic metabolites. However, current computational metabolism models are often unable to predict the specific metabolites formed by metabolism at certain sites. Identification of reaction type is a key step toward metabolite prediction. Phase I enzymes, which are responsible for the metabolism of more than 90% of FDA approved drugs, catalyze highly diverse types of reactions and produce metabolites with substantial structural variability. Without knowledge of potential metabolite structures, medicinal chemists cannot differentiate harmful metabolic transformations from beneficial ones. To address this shortcoming, we propose a system for simultaneously labeling sites of metabolism and reaction types, by classifying them into five key reaction classes: stable and unstable oxidations, dehydrogenation, hydrolysis, and reduction. These classes unambiguously identify 21 types of phase I reactions, which cover 92.3% of known reactions in our database. We used this labeling system to train a neural network model of phase I metabolism on a literature-derived data set encompassing 20 736 human phase I metabolic reactions. Our model, Rainbow XenoSite, was able to identify reaction-type specific sites of metabolism with a cross-validated accuracy of 97.1% area under the receiver operator curve. Rainbow XenoSite with five-color and combined output is available for use free and online through our secure server at http://swami.wustl.edu/xenosite/p/phase1_rainbow.

Terbinafine N-dealkylation pathways result in formation of 6,6-dimethyl-2-hepten-4-ynal (TBF-A), a reactive allylic aldehyde, that may initiate idiosyncratic drug-induced liver toxicity. Previously, we reported on the importance of CYP2C19 and 3A4 as major contributors to TBF-A formation. In this study, we expanded on those efforts to assess individual contributions of CYP1A2, 2B6, 2C8, 2C9, and 2D6 in terbinafine metabolism. The combined knowledge gained from these studies allowed us to scale the relative roles of the P450 isozymes in hepatic clearance of terbinafine including pathways leading to TBF-A, and hence, provide a foundation for assessing their significance in terbinafine-induced hepatotoxicity. We used in vitro terbinafine reactions with recombinant P450s to measure kinetics for multiple metabolic pathways and calculated contributions of all individual P450 isozymes to in vivo hepatic clearance for the average human adult. The findings confirmed that CYP3A4 was a major contributor (at least 30% total metabolism) to all three of the possible N-dealkylation pathways; however, CYP2C9, and not CYP2C19, played a critical role in terbinafine metabolism and even exceeded CYP3A4 contributions for terbinafine N-demethylation. A combination of their metabolic capacities accounted for at least 80% of the conversion of terbinafine to TBF-A, while CYP1A2, 2B6, 2C8, and 2D6 made minor contributions. Computational approaches provide a more rapid, less resource-intensive strategy for assessing metabolism, and thus, we additionally predicted terbinafine metabolism using deep neural network models for individual P450 isozymes. Cytochrome P450 isozyme models accurately predicted the likelihood for terbinafine N-demethylation, but overestimated the likelihood for a minor N-denaphthylation pathway. Moreover, the models were not able to differentiate the varying roles of the individual P450 isozymes for specific reactions with this particular drug. Taken together, the significance of CYP2C9 and 3A4 and to a lesser extent, CYP2C19, in terbinafine metabolism is consistent with reported drug interactions. This finding suggests that variations in individual P450 contributions due to other factors like polymorphisms may similarly contribute to terbinafine-related adverse health outcomes. Nevertheless, the impact of their metabolic capacities on formation of reactive TBF-A and consequent idiosyncratic hepatotoxicity will be mitigated by competing detoxification pathways, TBF-A decay, and TBF-A adduction to glutathione that remain understudied., (Copyright © 2019 The Authors. Published by Elsevier Inc. All rights reserved.)

A workshop on research gaps and opportunities for Precision Medicine in Pancreatic Disease was sponsored by the National Institute of Diabetes and Digestive Kidney Diseases on July 24, 2019, in Pittsburgh. The workshop included an overview lecture on precision medicine in cancer and 4 sessions: (1) general considerations for the application of bioinformatics and artificial intelligence; (2) omics, the combination of risk factors and biomarkers; (3) precision imaging; and (4) gaps, barriers, and needs to move from precision to personalized medicine for pancreatic disease. Current precision medicine approaches and tools were reviewed, and participants identified knowledge gaps and research needs that hinder bringing precision medicine to pancreatic diseases. Most critical were (a) multicenter efforts to collect large-scale patient data sets from multiple data streams in the context of environmental and social factors; (b) new information systems that can collect, annotate, and quantify data to inform disease mechanisms; (c) novel prospective clinical trial designs to test and improve therapies; and (d) a framework for measuring and assessing the value of proposed approaches to the health care system. With these advances, precision medicine can identify patients early in the course of their pancreatic disease and prevent progression to chronic or fatal illness.

Pediatric patients are at elevated risk of adverse drug reactions, and there is insufficient information on drug safety in children. Complicating risk assessment in children, there are numerous age-dependent changes in the absorption, distribution, metabolism, and elimination of drugs. A key contributor to age-dependent drug toxicity risk is the ontogeny of drug metabolism enzymes, the changes in both abundance and type throughout development from the fetal period through adulthood. Critically, these changes affect not only the overall clearance of drugs but also exposure to individual metabolites. In this study, we introduce time-embedding neural networks in order to model population-level variation in metabolism enzyme expression as a function of age. We use a time-embedding network to model the ontogeny of 23 drug metabolism enzymes. The time-embedding network recapitulates known demographic factors impacting 3A5 expression. The time-embedding network also effectively models the nonlinear dynamics of 2D6 expression, enabling a better fit to clinical data than prior work. In contrast, a standard neural network fails to model these features of 3A5 and 2D6 expression. Finally, we combine the time-embedding model of ontogeny with additional information to estimate age-dependent changes in reactive metabolite exposure. This simple approach identifies age-dependent changes in exposure to valproic acid and dextromethorphan metabolites and suggests potential mechanisms of valproic acid toxicity. This approach may help researchers evaluate the risk of drug toxicity in pediatric populations.

Lamisil (terbinafine) is an effective, widely prescribed antifungal drug that causes rare idiosyncratic hepatotoxicity. The proposed toxic mechanism involves a reactive metabolite, 6,6-dimethyl-2-hepten-4-ynal (TBF-A), formed through three N-dealkylation pathways. We were the first to characterize them using in vitro studies with human liver microsomes and modeling approaches, yet knowledge of the individual enzymes catalyzing reactions remained unknown. Herein, we employed experimental and computational tools to assess terbinafine metabolism by specific cytochrome P450 isozymes. In vitro inhibitor phenotyping studies revealed six isozymes were involved in one or more N-dealkylation pathways. CYP2C19 and 3A4 contributed to all pathways, and so, we targeted them for steady-state analyses with recombinant isozymes. N-Dealkylation yielding TBF-A directly was catalyzed by CYP2C19 and 3A4 similarly. Nevertheless, CYP2C19 was more efficient than CYP3A4 at N-demethylation and other steps leading to TBF-A. Unlike microsomal reactions, N-denaphthylation was surprisingly efficient for CYP2C19 and 3A4, which was validated by controls. CYP2C19 was the most efficient among all reactions. Nonetheless, CYP3A4 was more selective at steps leading to TBF-A, making it more effective in terbinafine bioactivation based on metabolic split ratios for competing pathways. Model predictions did not extrapolate to quantitative kinetic constants, yet some results for CYP3A4 and CYP2C19 agreed qualitatively with preferred reaction steps and pathways. Clinical data on drug interactions support the CYP3A4 role in terbinafine metabolism, while CYP2C19 remains understudied. Taken together, knowledge of P450s responsible for terbinafine metabolism and TBF-A formation provides a foundation for investigating and mitigating the impact of P450 variations in toxic risks posed to patients.

Purpose: Following automated variant calling, manual review of aligned read sequences is required to identify a high-quality list of somatic variants. Despite widespread use in analyzing sequence data, methods to standardize manual review have not been described, resulting in high inter- and intralab variability., Methods: This manual review standard operating procedure (SOP) consists of methods to annotate variants with four different calls and 19 tags. The calls indicate a reviewer's confidence in each variant and the tags indicate commonly observed sequencing patterns and artifacts that inform the manual review call. Four individuals were asked to classify variants prior to, and after, reading the SOP and accuracy was assessed by comparing reviewer calls with orthogonal validation sequencing., Results: After reading the SOP, average accuracy in somatic variant identification increased by 16.7% (p value = 0.0298) and average interreviewer agreement increased by 12.7% (p value < 0.001). Manual review conducted after reading the SOP did not significantly increase reviewer time., Conclusion: This SOP supports and enhances manual somatic variant detection by improving reviewer accuracy while reducing the interreviewer variability for variant calling and annotation.

Recent efforts to design personalized cancer immunotherapies use predicted neoantigens, but most neoantigen prediction strategies do not consider proximal (nearby) variants that alter the peptide sequence and may influence neoantigen binding. We evaluated somatic variants from 430 tumors to understand how proximal somatic and germline alterations change the neoantigenic peptide sequence and also affect neoantigen binding predictions. On average, 241 missense somatic variants were analyzed per sample. Of these somatic variants, 5% had one or more in-phase missense proximal variants. Without incorporating proximal variant correction for major histocompatibility complex class I neoantigen peptides, the overall false discovery rate (incorrect neoantigens predicted) and the false negative rate (strong-binding neoantigens missed) across peptides of lengths 8-11 were estimated as 0.069 (6.9%) and 0.026 (2.6%), respectively.

Transplantable kidneys are in very limited supply. Accurate viability assessment prior to transplantation could minimize organ discard. Rapid and accurate evaluation of intra-operative donor kidney biopsies is essential for determining which kidneys are eligible for transplantation. The criterion for accepting or rejecting donor kidneys relies heavily on pathologist determination of the percent of glomeruli (determined from a frozen section) that are normal and sclerotic. This percentage is a critical measurement that correlates with transplant outcome. Inter- and intra-observer variability in donor biopsy evaluation is, however, significant. An automated method for determination of percent global glomerulosclerosis could prove useful in decreasing evaluation variability, increasing throughput, and easing the burden on pathologists. Here, we describe the development of a deep learning model that identifies and classifies non-sclerosed and sclerosed glomeruli in whole-slide images of donor kidney frozen section biopsies. This model extends a convolutional neural network (CNN) pre-trained on a large database of digital images. The extended model, when trained on just 48 whole slide images, exhibits slide-level evaluation performance on par with expert renal pathologists. Encouragingly, the model's performance is robust to slide preparation artifacts associated with frozen section preparation. The model substantially outperforms a model trained on image patches of isolated glomeruli, in terms of both accuracy and speed. The methodology overcomes the technical challenge of applying a pretrained CNN bottleneck model to whole-slide image classification. The traditional patch-based approach, while exhibiting deceptively good performance classifying isolated patches, does not translate successfully to whole-slide image segmentation in this setting. As the first model reported that identifies and classifies normal and sclerotic glomeruli in frozen kidney sections, and thus the first model reported in the literature relevant to kidney transplantation, it may become an essential part of donor kidney biopsy evaluation in the clinical setting.

Cancer genomic analysis requires accurate identification of somatic variants in sequencing data. Manual review to refine somatic variant calls is required as a final step after automated processing. However, manual variant refinement is time-consuming, costly, poorly standardized, and non-reproducible. Here, we systematized and standardized somatic variant refinement using a machine learning approach. The final model incorporates 41,000 variants from 440 sequencing cases. This model accurately recapitulated manual refinement labels for three independent testing sets (13,579 variants) and accurately predicted somatic variants confirmed by orthogonal validation sequencing data (212,158 variants). The model improves on manual somatic refinement by reducing bias on calls otherwise subject to high inter-reviewer variability.

Lamisil (terbinafine) may cause idiosyncratic liver toxicity through a proposed toxicological mechanism involving the reactive metabolite 6,6-dimethyl-2-hepten-4-ynal (TBF-A). TBF-A toxicological relevance remains unclear due to a lack of identification of pathways leading to and competing with TBF-A formation. We resolved this knowledge gap by combining computational modeling and experimental kinetics of in vitro hepatic N-dealkylation of terbinafine. A deep learning model of N-dealkylation predicted a high probability for N-demethylation to yield desmethyl-terbinafine followed by N-dealkylation to TBF-A and marginal contributions from other possible pathways. We carried out steady-state kinetic experiments with pooled human liver microsomes that relied on development of labeling methods to expand metabolite characterization. Those efforts revealed high levels of TBF-A formation and first order decay during metabolic reactions; actual TBF-A levels would then reflect the balance between those processes as well as reflect the impact of stabilizing adduction with glutathione and other biological molecules. Modeling predictions and experimental studies agreed on the significance of N-demethylation and insignificance of N-denaphthylation in terbinafine metabolism, yet differed on importance of direct TBF-A formation. Under steady-state conditions, the direct pathway was the most important source of the reactive metabolite with a V max /K m of 4.0 pmol/min/mg protein/µM in contrast to model predictions. Nevertheless, previous studies show that therapeutic dosing leads to accumulation of desmethyl-terbinafine in plasma, which means that likely sources for TBF-A would draw from metabolism of both the major metabolite and parent drug based on our modeling and experimental studies. Through this combination of novel modeling and experimental approaches, we are the first to identify pathways leading to generation of TBF-A for assessing its role in idiosyncratic adverse drug interactions., (Copyright © 2018 Elsevier Inc. All rights reserved.)

Scientists rely on high-throughput screening tools to identify promising small-molecule compounds for the development of biochemical probes and drugs. This study focuses on the identification of promiscuous bioactive compounds, which are compounds that appear active in many high-throughput screening experiments against diverse targets but are often false-positives which may not be easily developed into successful probes. These compounds can exhibit bioactivity due to nonspecific, intractable mechanisms of action and/or by interference with specific assay technology readouts. Such "frequent hitters" are now commonly identified using substructure filters, including pan assay interference compounds (PAINS). Herein, we show that mechanistic modeling of small-molecule reactivity using deep learning can improve upon PAINS filters when modeling promiscuous bioactivity in PubChem assays. Without training on high-throughput screening data, a deep learning model of small-molecule reactivity achieves a sensitivity and specificity of 18.5% and 95.5%, respectively, in identifying promiscuous bioactive compounds. This performance is similar to PAINS filters, which achieve a sensitivity of 20.3% at the same specificity. Importantly, such reactivity modeling is complementary to PAINS filters. When PAINS filters and reactivity models are combined, the resulting model outperforms either method alone, achieving a sensitivity of 24% at the same specificity. However, as a probabilistic model, the sensitivity and specificity of the deep learning model can be tuned by adjusting the threshold. Moreover, for a subset of PAINS filters, this reactivity model can help discriminate between promiscuous and nonpromiscuous bioactive compounds even among compounds matching those filters. Critically, the reactivity model provides mechanistic hypotheses for assay interference by predicting the precise atoms involved in compound reactivity. Overall, our analysis suggests that deep learning approaches to modeling promiscuous compound bioactivity may provide a complementary approach to current methods for identifying promiscuous compounds.

Deep learning describes a class of machine learning algorithms that are capable of combining raw inputs into layers of intermediate features. These algorithms have recently shown impressive results across a variety of domains. Biology and medicine are data-rich disciplines, but the data are complex and often ill-understood. Hence, deep learning techniques may be particularly well suited to solve problems of these fields. We examine applications of deep learning to a variety of biomedical problems-patient classification, fundamental biological processes and treatment of patients-and discuss whether deep learning will be able to transform these tasks or if the biomedical sphere poses unique challenges. Following from an extensive literature review, we find that deep learning has yet to revolutionize biomedicine or definitively resolve any of the most pressing challenges in the field, but promising advances have been made on the prior state of the art. Even though improvements over previous baselines have been modest in general, the recent progress indicates that deep learning methods will provide valuable means for speeding up or aiding human investigation. Though progress has been made linking a specific neural network's prediction to input features, understanding how users should interpret these models to make testable hypotheses about the system under study remains an open challenge. Furthermore, the limited amount of labelled data for training presents problems in some domains, as do legal and privacy constraints on work with sensitive health records. Nonetheless, we foresee deep learning enabling changes at both bench and bedside with the potential to transform several areas of biology and medicine., (© 2018 The Authors.)

Cytochromes P450 (CYPs) oxidize alkylated amines commonly found in drugs and other biologically active molecules, cleaving them into an amine and an aldehyde. Metabolic studies usually neglect to report or investigate aldehydes, even though they can be toxic. It is assumed that they are efficiently detoxified into carboxylic acids and alcohols. Nevertheless, some aldehydes are reactive and escape detoxification pathways to cause adverse events by forming DNA and protein adducts. Herein, we modeled N-dealkylations that produce both amine and aldehyde metabolites and then predicted the reactivity of the aldehyde. This model used a deep learning approach previously developed by our group to predict other types of drug metabolism. In this study, we trained the model to predict N-dealkylation by human liver microsomes (HLM), finding that including isozyme-specific metabolism data alongside HLM data significantly improved results. The final HLM model accurately predicted the site of N-dealkylation within metabolized substrates (97% top-two and 94% area under the ROC curve). Next, we combined the metabolism, metabolite structure prediction, and previously published reactivity models into a bioactivation model. This combined model predicted the structure of the most likely reactive metabolite of a small validation set of drug-like molecules known to be bioactivated by N-dealkylation. Applying this model to approved and withdrawn medicines, we found that aldehyde metabolites produced from N-dealkylation may explain the hepatotoxicity of several drugs: indinavir, piperacillin, verapamil, and ziprasidone. Our results suggest that N-dealkylation may be an under-appreciated bioactivation pathway, especially in clinical contexts where aldehyde detoxification pathways are inhibited. Moreover, this is the first report of a bioactivation model constructed by combining a metabolism and reactivity model. These results raise hope that more comprehensive models of bioactivation are possible. The model developed in this study is available at http://swami.wustl.edu/xenosite/ .

A collection of new approaches to building and training neural networks, collectively referred to as deep learning, are attracting attention in theoretical chemistry. Several groups aim to replace computationally expensive ab initio quantum mechanics calculations with learned estimators. This raises questions about the representability of complex quantum chemical systems with neural networks. Can local-variable models efficiently approximate nonlocal quantum chemical features? Here, we find that convolutional architectures, those that only aggregate information locally, cannot efficiently represent aromaticity and conjugation in large systems. They cannot represent long-range nonlocality known to be important in quantum chemistry. This study uses aromatic and conjugated systems computed from molecule graphs, though reproducing quantum simulations is the ultimate goal. This task, by definition, is both computable and known to be important to chemistry. The failure of convolutional architectures on this focused task calls into question their use in modeling quantum mechanics. To remedy this heretofore unrecognized deficiency, we introduce a new architecture that propagates information back and forth in waves of nonlinear computation. This architecture is still a local-variable model, and it is both computationally and representationally efficient, processing molecules in sublinear time with far fewer parameters than convolutional networks. Wave-like propagation models aromatic and conjugated systems with high accuracy, and even models the impact of small structural changes on large molecules. This new architecture demonstrates that some nonlocal features of quantum chemistry can be efficiently represented in local variable models., Competing Interests: The authors declare no competing financial interest.

Motivation: Characterizing the binding specificities of transcription factors (TFs) is crucial to the study of gene expression regulation. Recently developed high-throughput experimental methods, including protein binding microarrays (PBM) and high-throughput SELEX (HT-SELEX), have enabled rapid measurements of the specificities for hundreds of TFs. However, few studies have developed efficient algorithms for estimating binding motifs based on HT-SELEX data. Also the simple method of constructing a position weight matrix (PWM) by comparing the frequency of the preferred sequence with single-nucleotide variants has the risk of generating motifs with higher information content than the true binding specificity., Results: We developed an algorithm called BEESEM that builds on a comprehensive biophysical model of protein-DNA interactions, which is trained using the expectation maximization method. BEESEM is capable of selecting the optimal motif length and calculating the confidence intervals of estimated parameters. By comparing BEESEM with the published motifs estimated using the same HT-SELEX data, we demonstrate that BEESEM provides significant improvements. We also evaluate several motif discovery algorithms on independent PBM and ChIP-seq data. BEESEM provides significantly better fits to in vitro data, but its performance is similar to some other methods on in vivo data under the criterion of the area under the receiver operating characteristic curve (AUROC). This highlights the limitations of the purely rank-based AUROC criterion. Using quantitative binding data to assess models, however, demonstrates that BEESEM improves on prior models., Availability and Implementation: Freely available on the web at http://stormo.wustl.edu/resources.html ., Contact: stormo@wustl.edu., Supplementary Information: Supplementary data are available at Bioinformatics online., (© The Author (2017). Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com)

Structural alerts are commonly used in drug discovery to identify molecules likely to form reactive metabolites and thereby become toxic. Unfortunately, as useful as structural alerts are, they do not effectively model if, when, and why metabolism renders safe molecules toxic. Toxicity due to a specific structural alert is highly conditional, depending on the metabolism of the alert, the reactivity of its metabolites, dosage, and competing detoxification pathways. A systems approach, which explicitly models these pathways, could more effectively assess the toxicity risk of drug candidates. In this study, we demonstrated that mathematical models of P450 metabolism can predict the context-specific probability that a structural alert will be bioactivated in a given molecule. This study focuses on the furan, phenol, nitroaromatic, and thiophene alerts. Each of these structural alerts can produce reactive metabolites through certain metabolic pathways but not always. We tested whether our metabolism modeling approach, XenoSite, can predict when a given molecule's alerts will be bioactivated. Specifically, we used models of epoxidation, quinone formation, reduction, and sulfur-oxidation to predict the bioactivation of furan-, phenol-, nitroaromatic-, and thiophene-containing drugs. Our models separated bioactivated and not-bioactivated furan-, phenol-, nitroaromatic-, and thiophene-containing drugs with AUC performances of 100%, 73%, 93%, and 88%, respectively. Metabolism models accurately predict whether alerts are bioactivated and thus serve as a practical approach to improve the interpretability and usefulness of structural alerts. We expect that this same computational approach can be extended to most other structural alerts and later integrated into toxicity risk models. This advance is one necessary step toward our long-term goal of building comprehensive metabolic models of bioactivation and detoxification to guide assessment and design of new therapeutic molecules.

Many adverse drug reactions are thought to be caused by electrophilically reactive drug metabolites that conjugate to nucleophilic sites within DNA and proteins, causing cancer or toxic immune responses. Quinone species, including quinone-imines, quinone-methides, and imine-methides, are electrophilic Michael acceptors that are often highly reactive and comprise over 40% of all known reactive metabolites. Quinone metabolites are created by cytochromes P450 and peroxidases. For example, cytochromes P450 oxidize acetaminophen to N-acetyl-p-benzoquinone imine, which is electrophilically reactive and covalently binds to nucleophilic sites within proteins. This reactive quinone metabolite elicits a toxic immune response when acetaminophen exceeds a safe dose. Using a deep learning approach, this study reports the first published method for predicting quinone formation: the formation of a quinone species by metabolic oxidation. We model both one- and two-step quinone formation, enabling accurate quinone formation predictions in nonobvious cases. We predict atom pairs that form quinones with an AUC accuracy of 97.6%, and we identify molecules that form quinones with 88.2% AUC. By modeling the formation of quinones, one of the most common types of reactive metabolites, our method provides a rapid screening tool for a key drug toxicity risk. The XenoSite quinone formation model is available at http://swami.wustl.edu/xenosite/p/quinone .

Motivation: Uridine diphosphate glucunosyltransferases (UGTs) metabolize 15% of FDA approved drugs. Lead optimization efforts benefit from knowing how candidate drugs are metabolized by UGTs. This paper describes a computational method for predicting sites of UGT-mediated metabolism on drug-like molecules., Results: XenoSite correctly predicts test molecule's sites of glucoronidation in the Top-1 or Top-2 predictions at a rate of 86 and 97%, respectively. In addition to predicting common sites of UGT conjugation, like hydroxyl groups, it can also accurately predict the glucoronidation of atypical sites, such as carbons. We also describe a simple heuristic model for predicting UGT-mediated sites of metabolism that performs nearly as well (with, respectively, 80 and 91% Top-1 and Top-2 accuracy), and can identify the most challenging molecules to predict on which to assess more complex models. Compared with prior studies, this model is more generally applicable, more accurate and simpler (not requiring expensive quantum modeling)., Availability and Implementation: The UGT metabolism predictor developed in this study is available at http://swami.wustl.edu/xenosite/p/ugt CONTACT: : swamidass@wustl.eduSupplementary information: Supplementary data are available at Bioinformatics online., (© The Author 2016. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.)

Methods are needed to reliably prioritize biologically active driver mutations over inactive passengers in high-throughput sequencing cancer data sets. We present ParsSNP, an unsupervised functional impact predictor that is guided by parsimony. ParsSNP uses an expectation-maximization framework to find mutations that explain tumor incidence broadly, without using predefined training labels that can introduce biases. We compare ParsSNP to five existing tools (CanDrA, CHASM, FATHMM Cancer, TransFIC, and Condel) across five distinct benchmarks. ParsSNP outperformed the existing tools in 24 of 25 comparisons. To investigate the real-world benefit of these improvements, we applied ParsSNP to an independent data set of 30 patients with diffuse-type gastric cancer. ParsSNP identified many known and likely driver mutations that other methods did not detect, including truncation mutations in known tumor suppressors and the recurrent driver substitution RHOA p.Tyr42Cys. In conclusion, ParsSNP uses an innovative, parsimony-based approach to prioritize cancer driver mutations and provides dramatic improvements over existing methods.

Most small-molecule drug candidates fail before entering the market, frequently because of unexpected toxicity. Often, toxicity is detected only late in drug development, because many types of toxicities, especially idiosyncratic adverse drug reactions (IADRs), are particularly hard to predict and detect. Moreover, drug-induced liver injury (DILI) is the most frequent reason drugs are withdrawn from the market and causes 50% of acute liver failure cases in the United States. A common mechanism often underlies many types of drug toxicities, including both DILI and IADRs. Drugs are bioactivated by drug-metabolizing enzymes into reactive metabolites, which then conjugate to sites in proteins or DNA to form adducts. DNA adducts are often mutagenic and may alter the reading and copying of genes and their regulatory elements, causing gene dysregulation and even triggering cancer. Similarly, protein adducts can disrupt their normal biological functions and induce harmful immune responses. Unfortunately, reactive metabolites are not reliably detected by experiments, and it is also expensive to test drug candidates for potential to form DNA or protein adducts during the early stages of drug development. In contrast, computational methods have the potential to quickly screen for covalent binding potential, thereby flagging problematic molecules and reducing the total number of necessary experiments. Here, we train a deep convolution neural network-the XenoSite reactivity model-using literature data to accurately predict both sites and probability of reactivity for molecules with glutathione, cyanide, protein, and DNA. On the site level, cross-validated predictions had area under the curve (AUC) performances of 89.8% for DNA and 94.4% for protein. Furthermore, the model separated molecules electrophilically reactive with DNA and protein from nonreactive molecules with cross-validated AUC performances of 78.7% and 79.8%, respectively. On both the site- and molecule-level, the model's performances significantly outperformed reactivity indices derived from quantum simulations that are reported in the literature. Moreover, we developed and applied a selectivity score to assess preferential reactions with the macromolecules as opposed to the common screening traps. For the entire data set of 2803 molecules, this approach yielded totals of 257 (9.2%) and 227 (8.1%) molecules predicted to be reactive only with DNA and protein, respectively, and hence those that would be missed by standard reactivity screening experiments. Site of reactivity data is an underutilized resource that can be used to not only predict if molecules are reactive, but also show where they might be modified to reduce toxicity while retaining efficacy. The XenoSite reactivity model is available at http://swami.wustl.edu/xenosite/p/reactivity.

Although epigenetic abnormalities have been described in Huntington's disease (HD), the causal epigenetic mechanisms driving neurodegeneration in HD cortex and striatum remain undefined. Using an epigenetic pathway-targeted drug screen, we report that inhibitors of DNA methyltransferases (DNMTs), decitabine and FdCyd, block mutant huntingtin (Htt)-induced toxicity in primary cortical and striatal neurons. In addition, knockdown of DNMT3A or DNMT1 protected neurons against mutant Htt-induced toxicity, together demonstrating a requirement for DNMTs in mutant Htt-triggered neuronal death and suggesting a neurodegenerative mechanism based on DNA methylation-mediated transcriptional repression. Inhibition of DNMTs in HD model primary cortical or striatal neurons restored the expression of several key genes, including Bdnf, an important neurotrophic factor implicated in HD. Accordingly, the Bdnf promoter exhibited aberrant cytosine methylation in mutant Htt-expressing cortical neurons. In vivo, pharmacological inhibition of DNMTs in HD mouse brains restored the mRNA levels of key striatal genes known to be downregulated in HD. Thus, disturbances in DNA methylation play a critical role in mutant Htt-induced neuronal dysfunction and death, raising the possibility that epigenetic strategies targeting abnormal DNA methylation may have therapeutic utility in HD.

A major cause of the paucity of new starting points for drug discovery is the lack of interaction between academia and industry. Much of the global resource in biology is present in universities, whereas the focus of medicinal chemistry is still largely within industry. Open source drug discovery, with sharing of information, is clearly a first step towards overcoming this gap. But the interface could especially be bridged through a scale-up of open sharing of physical compounds, which would accelerate the finding of new starting points for drug discovery. The Medicines for Malaria Venture Malaria Box is a collection of over 400 compounds representing families of structures identified in phenotypic screens of pharmaceutical and academic libraries against the Plasmodium falciparum malaria parasite. The set has now been distributed to almost 200 research groups globally in the last two years, with the only stipulation that information from the screens is deposited in the public domain. This paper reports for the first time on 236 screens that have been carried out against the Malaria Box and compares these results with 55 assays that were previously published, in a format that allows a meta-analysis of the combined dataset. The combined biochemical and cellular assays presented here suggest mechanisms of action for 135 (34%) of the compounds active in killing multiple life-cycle stages of the malaria parasite, including asexual blood, liver, gametocyte, gametes and insect ookinete stages. In addition, many compounds demonstrated activity against other pathogens, showing hits in assays with 16 protozoa, 7 helminths, 9 bacterial and mycobacterial species, the dengue fever mosquito vector, and the NCI60 human cancer cell line panel of 60 human tumor cell lines. Toxicological, pharmacokinetic and metabolic properties were collected on all the compounds, assisting in the selection of the most promising candidates for murine proof-of-concept experiments and medicinal chemistry programs. The data for all of these assays are presented and analyzed to show how outstanding leads for many indications can be selected. These results reveal the immense potential for translating the dispersed expertise in biological assays involving human pathogens into drug discovery starting points, by providing open access to new families of molecules, and emphasize how a small additional investment made to help acquire and distribute compounds, and sharing the data, can catalyze drug discovery for dozens of different indications. Another lesson is that when multiple screens from different groups are run on the same library, results can be integrated quickly to select the most valuable starting points for subsequent medicinal chemistry efforts.

Computational drug repositioning or repurposing is a promising and efficient tool for discovering new uses from existing drugs and holds the great potential for precision medicine in the age of big data. The explosive growth of large-scale genomic and phenotypic data, as well as data of small molecular compounds with granted regulatory approval, is enabling new developments for computational repositioning. To achieve the shortest path toward new drug indications, advanced data processing and analysis strategies are critical for making sense of these heterogeneous molecular measurements. In this review, we show recent advancements in the critical areas of computational drug repositioning from multiple aspects. First, we summarize available data sources and the corresponding computational repositioning strategies. Second, we characterize the commonly used computational techniques. Third, we discuss validation strategies for repositioning studies, including both computational and experimental methods. Finally, we highlight potential opportunities and use-cases, including a few target areas such as cancers. We conclude with a brief discussion of the remaining challenges in computational drug repositioning., (Published by Oxford University Press 2015. This work is written by US Government employees and is in the public domain in the US.)