BERT-based language model for accurate drug adverse event extraction from social media: implementation, evaluation, and contributions to pharmacovigilance practices.

Introduction: Social media platforms serve as a valuable resource for users to share health-related information, aiding in the monitoring of adverse events linked to medications and treatments in drug safety surveillance. However, extracting drug-related adverse events accurately and efficiently from social media poses challenges in both natural language processing research and the pharmacovigilance domain. Method: Recognizing the lack of detailed implementation and evaluation of Bidirectional Encoder Representations from Transformers (BERT)-based models for drug adverse event extraction on social media, we developed a BERT-based language model tailored to identifying drug adverse events in this context. Our model utilized publicly available labeled adverse event data from the ADE-Corpus-V2. Constructing the BERT-based model involved optimizing key hyperparameters, such as the number of training epochs, batch size, and learning rate. Through ten hold-out evaluations on ADE-Corpus-V2 data and external social media datasets, our model consistently demonstrated high accuracy in drug adverse event detection. Result: The hold-out evaluations resulted in average F1 scores of 0.8575, 0.9049, and 0.9813 for detecting words of adverse events, words in adverse events, and words not in adverse events, respectively. External validation using human-labeled adverse event tweets data from SMM4H further substantiated the effectiveness of our model, yielding F1 scores 0.8127, 0.8068, and 0.9790 for detecting words of adverse events, words in adverse events, and words not in adverse events, respectively. Discussion: This study not only showcases the effectiveness of BERT-based language models in accurately identifying drug-related adverse events in the dynamic landscape of social media data, but also addresses the need for the implementation of a comprehensive study design and evaluation. By doing so, we contribute to the advancement of pharmacovigilance practices and methodologies in the context of emerging information sources like social media.

Fingerprinting Interactions between Proteins and Ligands for Facilitating Machine Learning in Drug Discovery.

Molecular recognition is fundamental in biology, underpinning intricate processes through specific protein-ligand interactions. This understanding is pivotal in drug discovery, yet traditional experimental methods face limitations in exploring the vast chemical space. Computational approaches, notably quantitative structure-activity/property relationship analysis, have gained prominence. Molecular fingerprints encode molecular structures and serve as property profiles, which are essential in drug discovery. While two-dimensional (2D) fingerprints are commonly used, three-dimensional (3D) structural interaction fingerprints offer enhanced structural features specific to target proteins. Machine learning models trained on interaction fingerprints enable precise binding prediction. Recent focus has shifted to structure-based predictive modeling, with machine-learning scoring functions excelling due to feature engineering guided by key interactions. Notably, 3D interaction fingerprints are gaining ground due to their robustness. Various structural interaction fingerprints have been developed and used in drug discovery, each with unique capabilities. This review recapitulates the developed structural interaction fingerprints and provides two case studies to illustrate the power of interaction fingerprint-driven machine learning. The first elucidates structure-activity relationships in ß2 adrenoceptor ligands, demonstrating the ability to differentiate agonists and antagonists. The second employs a retrosynthesis-based pre-trained molecular representation to predict protein-ligand dissociation rates, offering insights into binding kinetics. Despite remarkable progress, challenges persist in interpreting complex machine learning models built on 3D fingerprints, emphasizing the need for strategies to make predictions interpretable. Binding site plasticity and induced fit effects pose additional complexities. Interaction fingerprints are promising but require continued research to harness their full potential.

Review of machine learning and deep learning models for toxicity prediction.

The ever-increasing number of chemicals has raised public concerns due to their adverse effects on human health and the environment. To protect public health and the environment, it is critical to assess the toxicity of these chemicals. Traditional in vitro and in vivo toxicity assays are complicated, costly, and time-consuming and may face ethical issues. These constraints raise the need for alternative methods for assessing the toxicity of chemicals. Recently, due to the advancement of machine learning algorithms and the increase in computational power, many toxicity prediction models have been developed using various machine learning and deep learning algorithms such as support vector machine, random forest, k-nearest neighbors, ensemble learning, and deep neural network. This review summarizes the machine learning- and deep learning-based toxicity prediction models developed in recent years. Support vector machine and random forest are the most popular machine learning algorithms, and hepatotoxicity, cardiotoxicity, and carcinogenicity are the frequently modeled toxicity endpoints in predictive toxicology. It is known that datasets impact model performance. The quality of datasets used in the development of toxicity prediction models using machine learning and deep learning is vital to the performance of the developed models. The different toxicity assignments for the same chemicals among different datasets of the same type of toxicity have been observed, indicating benchmarking datasets is needed for developing reliable toxicity prediction models using machine learning and deep learning algorithms. This review provides insights into current machine learning models in predictive toxicology, which are expected to promote the development and application of toxicity prediction models in the future.

Machine learning and deep learning for brain tumor MRI image segmentation.

Brain tumors are often fatal. Therefore, accurate brain tumor image segmentation is critical for the diagnosis, treatment, and monitoring of patients with these tumors. Magnetic resonance imaging (MRI) is a commonly used imaging technique for capturing brain images. Both machine learning and deep learning techniques are popular in analyzing MRI images. This article reviews some commonly used machine learning and deep learning techniques for brain tumor MRI image segmentation. The limitations and advantages of the reviewed machine learning and deep learning methods are discussed. Even though each of these methods has a well-established status in their individual domains, the combination of two or more techniques is currently an emerging trend.

Developing a SARS-CoV-2 main protease binding prediction random forest model for drug repurposing for COVID-19 treatment.

The coronavirus disease 2019 (COVID-19) global pandemic resulted in millions of people becoming infected with the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus and close to seven million deaths worldwide. It is essential to further explore and design effective COVID-19 treatment drugs that target the main protease of SARS-CoV-2, a major target for COVID-19 drugs. In this study, machine learning was applied for predicting the SARS-CoV-2 main protease binding of Food and Drug Administration (FDA)-approved drugs to assist in the identification of potential repurposing candidates for COVID-19 treatment. Ligands bound to the SARS-CoV-2 main protease in the Protein Data Bank and compounds experimentally tested in SARS-CoV-2 main protease binding assays in the literature were curated. These chemicals were divided into training (516 chemicals) and testing (360 chemicals) data sets. To identify SARS-CoV-2 main protease binders as potential candidates for repurposing to treat COVID-19, 1188 FDA-approved drugs from the Liver Toxicity Knowledge Base were obtained. A random forest algorithm was used for constructing predictive models based on molecular descriptors calculated using Mold2 software. Model performance was evaluated using 100 iterations of fivefold cross-validations which resulted in 78.8% balanced accuracy. The random forest model that was constructed from the whole training dataset was used to predict SARS-CoV-2 main protease binding on the testing set and the FDA-approved drugs. Model applicability domain and prediction confidence on drugs predicted as the main protease binders discovered 10 FDA-approved drugs as potential candidates for repurposing to treat COVID-19. Our results demonstrate that machine learning is an efficient method for drug repurposing and, thus, may accelerate drug development targeting SARS-CoV-2.

Analyzing 3D structures of the SARS-CoV-2 main protease reveals structural features of ligand binding for COVID-19 drug discovery.

The severe acute respiratory syndrome-coronavirus 2 (SARS-CoV-2) main protease has an essential role in viral replication and has become a major target for coronavirus 2019 (COVID-19) drug development. Various inhibitors have been discovered or designed to bind to the main protease. The availability of more than 550 3D structures of the main protease provides a wealth of structural details on the main protease in both ligand-free and ligand-bound states. Therefore, we examined these structures to ascertain the structural features for the role of the main protease in the cleavage of polyproteins, the alternative conformations during main protease maturation, and ligand interactions in the main protease. The structural features unearthed could promote the development of COVID-19 drugs targeting the SARS-CoV-2 main protease.

Development of a primate model to evaluate the effects of ketamine and surgical stress on the neonatal brain.

With advances in pediatric and obstetric surgery, pediatric patients are subject to complex procedures under general anesthesia. The effects of anesthetic exposure on the developing brain may be confounded by several factors including pre-existing disorders and surgery-induced stress. Ketamine, a noncompetitive N-methyl-d-aspartate (NMDA) receptor antagonist, is routinely used as a pediatric general anesthetic. However, controversy remains about whether ketamine exposure may be neuroprotective or induce neuronal degeneration in the developing brain. Here, we report the effects of ketamine exposure on the neonatal nonhuman primate brain under surgical stress. Eight neonatal rhesus monkeys (postnatal days 5-7) were randomly assigned to each of two groups: Group A (n = 4) received 2 mg/kg ketamine via intravenous bolus prior to surgery and a 0.5 mg/kg/h ketamine infusion during surgery in the presence of a standardized pediatric anesthetic regimen; Group B (n = 4) received volumes of normal saline equivalent to those of ketamine given to Group A animals prior to and during surgery, also in the presence of a standardized pediatric anesthetic regimen. Under anesthesia, the surgery consisted of a thoracotomy followed by closing the pleural space and tissue in layers using standard surgical techniques. Vital signs were monitored to be within normal ranges throughout anesthesia. Elevated levels of cytokines interleukin (IL)-8, IL-15, monocyte chemoattractant protein-1 (MCP-1), and macrophage inflammatory protein (MIP)-1ß at 6 and 24 h after surgery were detected in ketamine-exposed animals. Fluoro-Jade C staining revealed significantly higher neuronal degeneration in the frontal cortex of ketamine-exposed animals, compared with control animals. Intravenous ketamine administration prior to and throughout surgery in a clinically relevant neonatal primate model appears to elevate cytokine levels and increase neuronal degeneration. Consistent with previous data on the effects of ketamine on the developing brain, the results from the current randomized controlled study in neonatal monkeys undergoing simulated surgery show that ketamine does not provide neuroprotective or anti-inflammatory effects.

Three-Dimensional Structural Insights Have Revealed the Distinct Binding Interactions of Agonists, Partial Agonists, and Antagonists with the µ Opioid Receptor.

The United States is experiencing the most profound and devastating opioid crisis in history, with the number of deaths involving opioids, including prescription and illegal opioids, continuing to climb over the past two decades. This severe public health issue is difficult to combat as opioids remain a crucial treatment for pain, and at the same time, they are also highly addictive. Opioids act on the opioid receptor, which in turn activates its downstream signaling pathway that eventually leads to an analgesic effect. Among the four types of opioid receptors, the µ subtype is primarily responsible for the analgesic cascade. This review describes available 3D structures of the µ opioid receptor in the protein data bank and provides structural insights for the binding of agonists and antagonists to the receptor. Comparative analysis on the atomic details of the binding site in these structures was conducted and distinct binding interactions for agonists, partial agonists, and antagonists were observed. The findings in this article deepen our understanding of the ligand binding activity and shed some light on the development of novel opioid analgesics which may improve the risk benefit balance of existing opioids.

Machine learning models for rat multigeneration reproductive toxicity prediction.

Reproductive toxicity is one of the prominent endpoints in the risk assessment of environmental and industrial chemicals. Due to the complexity of the reproductive system, traditional reproductive toxicity testing in animals, especially guideline multigeneration reproductive toxicity studies, take a long time and are expensive. Therefore, machine learning, as a promising alternative approach, should be considered when evaluating the reproductive toxicity of chemicals. We curated rat multigeneration reproductive toxicity testing data of 275 chemicals from ToxRefDB (Toxicity Reference Database) and developed predictive models using seven machine learning algorithms (decision tree, decision forest, random forest, k-nearest neighbors, support vector machine, linear discriminant analysis, and logistic regression). A consensus model was built based on the seven individual models. An external validation set was curated from the COSMOS database and the literature. The performances of individual and consensus models were evaluated using 500 iterations of 5-fold cross-validations and the external validation data set. The balanced accuracy of the models ranged from 58% to 65% in the 5-fold cross-validations and 45%-61% in the external validations. Prediction confidence analysis was conducted to provide additional information for more appropriate applications of the developed models. The impact of our findings is in increasing confidence in machine learning models. We demonstrate the importance of using consensus models for harnessing the benefits of multiple machine learning models (i.e., using redundant systems to check validity of outcomes). While we continue to build upon the models to better characterize weak toxicants, there is current utility in saving resources by being able to screen out strong reproductive toxicants before investing in vivo testing. The modeling approach (machine learning models) is offered for assessing the rat multigeneration reproductive toxicity of chemicals. Our results suggest that machine learning may be a promising alternative approach to evaluate the potential reproductive toxicity of chemicals.

Machine Learning Models for Predicting Liver Toxicity.

Liver toxicity is a major adverse drug reaction that accounts for drug failure in clinical trials and withdrawal from the market. Therefore, predicting potential liver toxicity at an early stage in drug discovery is crucial to reduce costs and the potential for drug failure. However, current in vivo animal toxicity testing is very expensive and time consuming. As an alternative approach, various machine learning models have been developed to predict potential liver toxicity in humans. This chapter reviews current advances in the development and application of machine learning models for prediction of potential liver toxicity in humans and discusses possible improvements to liver toxicity prediction.

Application of Nonhuman Primate Models in the Studies of Pediatric Anesthesia Neurotoxicity.

Numerous animal models have been used to study developmental neurotoxicity associated with short-term or prolonged exposure of common general anesthetics at clinically relevant concentrations. Pediatric anesthesia models using the nonhuman primate (NHP) may more accurately reflect the human condition because of their phylogenetic similarity to humans with regard to reproduction, development, neuroanatomy, and cognition. Although they are not as widely used as other animal models, the contribution of NHP models in the study of anesthetic-induced developmental neurotoxicity has been essential. In this review, we discuss how neonatal NHP animals have been used for modeling pediatric anesthetic exposure; how NHPs have addressed key data gaps and application of the NHP model for the studies of general anesthetic-induced developmental neurotoxicity. The appropriate application and evaluation of the NHP model in the study of general anesthetic-induced developmental neurotoxicity have played a key role in enhancing the understanding and awareness of the potential neurotoxicity associated with pediatric general anesthetics.

Machine Learning Models for Predicting Cytotoxicity of Nanomaterials.

The wide application of nanomaterials in consumer and medical products has raised concerns about their potential adverse effects on human health. Thus, more and more biological assessments regarding the toxicity of nanomaterials have been performed. However, the different ways the evaluations were performed, such as the utilized assays, cell lines, and the differences of the produced nanoparticles, make it difficult for scientists to analyze and effectively compare toxicities of nanomaterials. Fortunately, machine learning has emerged as a powerful tool for the prediction of nanotoxicity based on the available data. Among different types of toxicity assessments, nanomaterial cytotoxicity was the focus here because of the high sensitivity of cytotoxicity assessment to different treatments without the need for complicated and time-consuming procedures. In this review, we summarized recent studies that focused on the development of machine learning models for prediction of cytotoxicity of nanomaterials. The goal was to provide insight into predicting potential nanomaterial toxicity and promoting the development of safe nanomaterials.

Elucidation of Agonist and Antagonist Dynamic Binding Patterns in ER-α by Integration of Molecular Docking, Molecular Dynamics Simulations and Quantum Mechanical Calculations.

Estrogen receptor alpha (ERα) is a ligand-dependent transcriptional factor in the nuclear receptor superfamily. Many structures of ERα bound with agonists and antagonists have been determined. However, the dynamic binding patterns of agonists and antagonists in the binding site of ERα remains unclear. Therefore, we performed molecular docking, molecular dynamics (MD) simulations, and quantum mechanical calculations to elucidate agonist and antagonist dynamic binding patterns in ERα. 17ß-estradiol (E2) and 4-hydroxytamoxifen (OHT) were docked in the ligand binding pockets of the agonist and antagonist bound ERα. The best complex conformations from molecular docking were subjected to 100 nanosecond MD simulations. Hierarchical clustering was conducted to group the structures in the trajectory from MD simulations. The representative structure from each cluster was selected to calculate the binding interaction energy value for elucidation of the dynamic binding patterns of agonists and antagonists in the binding site of ERα. The binding interaction energy analysis revealed that OHT binds ERα more tightly in the antagonist conformer, while E2 prefers the agonist conformer. The results may help identify ERα antagonists as drug candidates and facilitate risk assessment of chemicals through ER-mediated responses.

Nanomaterial Databases: Data Sources for Promoting Design and Risk Assessment of Nanomaterials.

Nanomaterials have drawn increasing attention due to their tunable and enhanced physicochemical and biological performance compared to their conventional bulk materials. Owing to the rapid expansion of the nano-industry, large amounts of data regarding the synthesis, physicochemical properties, and bioactivities of nanomaterials have been generated. These data are a great asset to the scientific community. However, the data are on diverse aspects of nanomaterials and in different sources and formats. To help utilize these data, various databases on specific information of nanomaterials such as physicochemical characterization, biomedicine, and nano-safety have been developed and made available online. Understanding the structure, function, and available data in these databases is needed for scientists to select appropriate databases and retrieve specific information for research on nanomaterials. However, to our knowledge, there is no study to systematically compare these databases to facilitate their utilization in the field of nanomaterials. Therefore, we reviewed and compared eight widely used databases of nanomaterials, aiming to provide the nanoscience community with valuable information about the specific content and function of these databases. We also discuss the pros and cons of these databases, thus enabling more efficient and convenient utilization.

Informing selection of drugs for COVID-19 treatment through adverse events analysis.

Coronavirus disease 2019 (COVID-19) is an ongoing pandemic and there is an urgent need for safe and effective drugs for COVID-19 treatment. Since developing a new drug is time consuming, many approved or investigational drugs have been repurposed for COVID-19 treatment in clinical trials. Therefore, selection of safe drugs for COVID-19 patients is vital for combating this pandemic. Our goal was to evaluate the safety concerns of drugs by analyzing adverse events reported in post-market surveillance. We collected 296 drugs that have been evaluated in clinical trials for COVID-19 and identified 28,597,464 associated adverse events at the system organ classes (SOCs) level in the FDA adverse events report systems (FAERS). We calculated Z-scores of SOCs that statistically quantify the relative frequency of adverse events of drugs in FAERS to quantitatively measure safety concerns for the drugs. Analyzing the Z-scores revealed that these drugs are associated with different significantly frequent adverse events. Our results suggest that this safety concern metric may serve as a tool to inform selection of drugs with favorable safety profiles for COVID-19 patients in clinical practices. Caution is advised when administering drugs with high Z-scores to patients who are vulnerable to associated adverse events.

Identification of Epidemiological Traits by Analysis of SARS-CoV-2 Sequences.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused the ongoing global COVID-19 pandemic that began in late December 2019. The rapid spread of SARS-CoV-2 is primarily due to person-to-person transmission. To understand the epidemiological traits of SARS-CoV-2 transmission, we conducted phylogenetic analysis on genome sequences from >54K SARS-CoV-2 cases obtained from two public databases. Hierarchical clustering analysis on geographic patterns in the resulting phylogenetic trees revealed a co-expansion tendency of the virus among neighboring countries with diverse sources and transmission routes for SARS-CoV-2. Pairwise sequence similarity analysis demonstrated that SARS-CoV-2 is transmitted locally and evolves during transmission. However, no significant differences were seen among SARS-CoV-2 genomes grouped by host age or sex. Here, our identified epidemiological traits provide information to better prevent transmission of SARS-CoV-2 and to facilitate the development of effective vaccines and therapeutics against the virus.

Neuroprotective Effects of Carnitine and Its Potential Application to Ameliorate Neurotoxicity.

Carnitine is an essential metabolite that is absorbed from the diet and synthesized in the kidney, liver, and brain. It ferries fatty acids across the mitochondrial membrane to undergo ß-oxidation. Carnitine has been studied as a therapy or protective agent for many neurological diseases and neurotoxicity (e.g., prolonged anesthetic exposure-induced developmental neurotoxicity in preclinical models). Preclinical and clinical data support the notion that carnitine or acetyl carnitine may improve a patient's quality of life through increased mitochondrial respiration, release of neurotransmitters, and global gene expression changes, showing the potential of carnitine beyond its approved use to treat primary and secondary carnitine deficiency. In this review, we summarize the beneficial effects of carnitine or acetyl carnitine on the central nervous system, highlighting protective effects against neurotoxicity-induced damage caused by various chemicals and encouraging a thorough evaluation of carnitine use as a therapy for patients suffering from neurotoxicant exposure.

Application of microRNA profiling to understand sevoflurane-induced adverse effects on developing monkey brain.

We have described that prolonged sevoflurane exposure at a clinically-relevant concentration of 2.5 % caused neuronal cell death in the developing monkey brain. Postnatal day 5 or 6 rhesus monkeys (n = 3) were exposed to 2.5 % sevoflurane for 8 h. Monkeys kept at environmental conditions in the procedure room served as controls (n = 3). Brain tissues were harvested four hours after sevoflurane exposure for histological analysis, and RNA or protein extraction. MicroRNA (miRNA) profiling on the frontal cortex of monkey brains was performed using next-generation sequencing. 417 miRNAs were identified in the frontal cortex, where most neuronal cell death was observed. 7 miRNAs were differentially expressed in frontal cortex after sevoflurane exposure. Five of these were expressed at significantly lower levels than controls and the other two miRNAs were expressed significantly higher. These differentially expressed miRNAs (DEMs) were then loaded to the Ingenuity Pathway Analysis database for pathway analysis, in which targeting information was available for 5 DEMs. The 5 DEMs target 2,919 mRNAs which are involved in pathways that contribute to various cellular functions. Of note, 78 genes that are related to axon guidance signaling were targeted, suggesting that development of the neural circuit may be affected after sevoflurane exposure. Such changes may have long-term effects on brain development and function. These findings are supplementary to our previous observations and provide more evidence for better understanding the adverse effects of sevoflurane on the developing brain after an 8 -h exposure.

Elucidating Interactions Between SARS-CoV-2 Trimeric Spike Protein and ACE2 Using Homology Modeling and Molecular Dynamics Simulations.

Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) causes coronavirus disease 2019 (COVID-19). As of October 21, 2020, more than 41.4 million confirmed cases and 1.1 million deaths have been reported. Thus, it is immensely important to develop drugs and vaccines to combat COVID-19. The spike protein present on the outer surface of the virion plays a major role in viral infection by binding to receptor proteins present on the outer membrane of host cells, triggering membrane fusion and internalization, which enables release of viral ssRNA into the host cell. Understanding the interactions between the SARS-CoV-2 trimeric spike protein and its host cell receptor protein, angiotensin converting enzyme 2 (ACE2), is important for developing drugs and vaccines to prevent and treat COVID-19. Several crystal structures of partial and mutant SARS-CoV-2 spike proteins have been reported; however, an atomistic structure of the wild-type SARS-CoV-2 trimeric spike protein complexed with ACE2 is not yet available. Therefore, in our study, homology modeling was used to build the trimeric form of the spike protein complexed with human ACE2, followed by all-atom molecular dynamics simulations to elucidate interactions at the interface between the spike protein and ACE2. Molecular Mechanics Poisson-Boltzmann Surface Area (MMPBSA) and in silico alanine scanning were employed to characterize the interacting residues at the interface. Twenty interacting residues in the spike protein were identified that are likely to be responsible for tightly binding to ACE2, of which five residues (Val445, Thr478, Gly485, Phe490, and Ser494) were not reported in the crystal structure of the truncated spike protein receptor binding domain (RBD) complexed with ACE2. These data indicate that the interactions between ACE2 and the tertiary structure of the full-length spike protein trimer are different from those between ACE2 and the truncated monomer of the spike protein RBD. These findings could facilitate the development of drugs and vaccines to prevent SARS-CoV-2 infection and combat COVID-19.