PK-DB: pharmacokinetics database for individualized and stratified computational modeling

Utility of life stage-specific chemical risk assessments based on New Approach Methodologies (NAMs)

The safety assessments for chemicals targeted for use or expected to be exposed to specific life stages, including infancy, childhood, pregnancy and lactation, and geriatrics, need to account for extrapolation of data from healthy adults to these populations to assess their human health risk. However, often adequate and relevant toxicity or pharmacokinetic (PK) data of chemicals in specific life stages are not available. For such chemicals, New Approach Methodologies (NAMs), such as physiologically based pharmacokinetic (PBPK) modeling, biologically based dose response (BBDR) modeling, in vitro to in vivo extrapolation (IVIVE), etc. can be used to understand the variability of exposure and effects of chemicals in specific life stages and assess their associated risk. A life stage specific PBPK model incorporates the physiological and biochemical changes associated with each life stage and simulates their impact on the absorption, distribution, metabolism, and elimination (ADME) of these chemicals. In our review, we summarize the parameterization of life stage models based on New Approach Methodologies (NAMs) and discuss case studies that highlight the utility of a life stage based PBPK modeling for risk assessment. In addition, we discuss the utility of artificial intelligence (AI)/machine learning (ML) and other computational models, such as those based on in vitro data, as tools for estimation of relevant physiological or physicochemical parameters and selection of model. We also discuss existing gaps in the available toxicological datasets and current challenges that need to be overcome to expand the utility of NAMs for life stage-specific chemical risk assessment.

Exploring pectin-casein micelles as novel carriers for oral drug delivery of artesunate in the treatment of systemic lupus erythematosus

The oral route of administration is considered the optimal choice for treating chronic diseases due to its convenience and non-invasiveness, which can help prevent physical and mental harm to patients undergoing long-term treatment. However, challenges such as safety, gastrointestinal stability, and bioavailability of oral drugs often limit their effectiveness. Natural biomacromolecule micelles, known for their safety, stability, biocompatibility, and diverse functions, have emerged as promising carriers for oral treatment of chronic diseases like systemic lupus erythematosus (SLE) with fat-soluble drugs. This study introduces an innovative approach by developing an oral delivery system using chemically synthesized natural biomacromolecules to load artesunate for treating SLE. By synthesizing amphiphilic polymer micelles from pectin and casein through a carbodiimide reaction, a more stable structure is achieved. The hydrophobic core of these micelles encapsulates artesunate, resulting in the formation of an oral delivery system (PC-AS) with several advantages, including high drug loading and encapsulation efficiency, small particle size, negative potential, strong stability in the gastrointestinal tract, low toxicity and side effects, strong adhesion in the small intestine, and high bioavailability. These advantages facilitate efficient absorption of artesunate in the gastrointestinal tract, leading to improved bioavailability and effective alleviation of SLE-like symptoms in MRL/lpr mice. By utilizing chemically synthesized natural macromolecular micelles for delivering artesunate in the treatment of SLE, this study overcomes the oral barriers associated with the original drug and presents a novel solution for the long-term oral treatment of chronic diseases.

Translational bioinformatics approach to combat cardiovascular disease and cancers

Bioinformatics is an interconnected subject of science dealing with diverse fields including biology, chemistry, physics, statistics, mathematics, and computer science as the key fields to answer complicated physiological problems. Key intention of bioinformatics is to store, analyze, organize, and retrieve essential information about genome, proteome, transcriptome, metabolome, as well as organisms to investigate the biological system along with its dynamics, if any. The outcome of bioinformatics depends on the type, quantity, and quality of the raw data provided and the algorithm employed to analyze the same. Despite several approved medicines available, cardiovascular disorders (CVDs) and cancers comprises of the two leading causes of human deaths. Understanding the unknown facts of both these non-communicable disorders is inevitable to discover new pathways, find new drug targets, and eventually newer drugs to combat them successfully. Since, all these goals involve complex investigation and handling of various types of macro- and small- molecules of the human body, bioinformatics plays a key role in such processes. Results from such investigation has direct human application and thus we call this filed as translational bioinformatics. Current book chapter thus deals with diverse scope and applications of this translational bioinformatics to find cure, diagnosis, and understanding the mechanisms of CVDs and cancers. Developing complex yet small or long algorithms to address such problems is very common in translational bioinformatics. Structure-based drug discovery or AI-guided invention of novel antibodies that too with super-high accuracy, speed, and involvement of considerably low amount of investment are some of the astonishing features of the translational bioinformatics and its applications in the fields of CVDs and cancers.

A data-driven machine learning approach for discovering potent LasR inhibitors

The rampant spread of multidrug-resistant Pseudomonas aeruginosa strains severely threatens global health. This severity is compounded against the backdrop of a stagnating antibiotics development pipeline. Moreover, with many promising therapeutics falling short of expectations in clinical trials, targeting the las quorum sensing (QS) system remains an attractive therapeutic strategy to combat P. aeruginosa infection. Thus, our primary goal was to develop a drug prediction algorithm using machine learning to identify potent LasR inhibitors. In this work, we demonstrated using a Multilayer Perceptron (MLP) algorithm boosted with AdaBoostM1 to discriminate between active and inactive LasR inhibitors. The optimal model performance was evaluated using 5-fold cross-validation and test sets. Our best model achieved a 90.7% accuracy in distinguishing active from inactive LasR inhibitors, an area under the Receiver Operating Characteristic Curve value of 0.95, and a Matthews correlation coefficient value of 0.81 when evaluated using test sets. Subsequently, we deployed the model against the Enamine database. The top-ranked compounds were further evaluated for their target engagement activity using molecular docking studies, Molecular Dynamics simulations, MM-GBSA analysis, and Free Energy Landscape analysis. Our data indicate that several of our chosen top hits showed better ligand-binding affinities than naringenin, a competitive LasR inhibitor. Among the six top hits, five of these compounds were predicted to be LasR inhibitors that could be used to treat P. aeruginosa-associated infections. To our knowledge, this study provides the first assessment of using an MLP-based QSAR model for discovering potent LasR inhibitors to attenuate P. aeruginosa infections.

A pathway model of glucose-stimulated insulin secretion in the pancreatic β-cell

The pancreas plays a critical role in maintaining glucose homeostasis through the secretion of hormones from the islets of Langerhans. Glucose-stimulated insulin secretion (GSIS) by the pancreatic β-cell is the main mechanism for reducing elevated plasma glucose. Here we present a systematic modeling workflow for the development of kinetic pathway models using the Systems Biology Markup Language (SBML). Steps include retrieval of information from databases, curation of experimental and clinical data for model calibration and validation, integration of heterogeneous data including absolute and relative measurements, unit normalization, data normalization, and model annotation. An important factor was the reproducibility and exchangeability of the model, which allowed the use of various existing tools. The workflow was applied to construct a novel data-driven kinetic model of GSIS in the pancreatic β-cell based on experimental and clinical data from 39 studies spanning 50 years of pancreatic, islet, and β-cell research in humans, rats, mice, and cell lines. The model consists of detailed glycolysis and phenomenological equations for insulin secretion coupled to cellular energy state, ATP dynamics and (ATP/ADP ratio). Key findings of our work are that in GSIS there is a glucose-dependent increase in almost all intermediates of glycolysis. This increase in glycolytic metabolites is accompanied by an increase in energy metabolites, especially ATP and NADH. One of the few decreasing metabolites is ADP, which, in combination with the increase in ATP, results in a large increase in ATP/ADP ratios in the β-cell with increasing glucose. Insulin secretion is dependent on ATP/ADP, resulting in glucose-stimulated insulin secretion. The observed glucose-dependent increase in glycolytic intermediates and the resulting change in ATP/ADP ratios and insulin secretion is a robust phenomenon observed across data sets, experimental systems and species. Model predictions of the glucose-dependent response of glycolytic intermediates and biphasic insulin secretion are in good agreement with experimental measurements. Our model predicts that factors affecting ATP consumption, ATP formation, hexokinase, phosphofructokinase, and ATP/ADP-dependent insulin secretion have a major effect on GSIS. In conclusion, we have developed and applied a systematic modeling workflow for pathway models that allowed us to gain insight into key mechanisms in GSIS in the pancreatic β-cell.

Multiple sampling schemes and deep learning improve active learning performance in drug-drug interaction information retrieval analysis from the literature

BACKGROUND

Drug-drug interaction (DDI) information retrieval (IR) is an important natural language process (NLP) task from the PubMed literature. For the first time, active learning (AL) is studied in DDI IR analysis. DDI IR analysis from PubMed abstracts faces the challenges of relatively small positive DDI samples among overwhelmingly large negative samples. Random negative sampling and positive sampling are purposely designed to improve the efficiency of AL analysis. The consistency of random negative sampling and positive sampling is shown in the paper.

RESULTS

PubMed abstracts are divided into two pools. Screened pool contains all abstracts that pass the DDI keywords query in PubMed, while unscreened pool includes all the other abstracts. At a prespecified recall rate of 0.95, DDI IR analysis precision is evaluated and compared. In screened pool IR analysis using supporting vector machine (SVM), similarity sampling plus uncertainty sampling improves the precision over uncertainty sampling, from 0.89 to 0.92 respectively. In the unscreened pool IR analysis, the integrated random negative sampling, positive sampling, and similarity sampling improve the precision over uncertainty sampling along, from 0.72 to 0.81 respectively. When we change the SVM to a deep learning method, all sampling schemes consistently improve DDI AL analysis in both screened pool and unscreened pool. Deep learning has significant improvement of precision over SVM, 0.96 vs. 0.92 in screened pool, and 0.90 vs. 0.81 in the unscreened pool, respectively.

CONCLUSIONS

By integrating various sampling schemes and deep learning algorithms into AL, the DDI IR analysis from literature is significantly improved. The random negative sampling and positive sampling are highly effective methods in improving AL analysis where the positive and negative samples are extremely imbalanced.

An Automated Customizable Live Web Crawler for Curation of Comparative Pharmacokinetic Data: An Intelligent Compilation of Research-Based Comprehensive Article Repository

Data curation has significant research implications irrespective of application areas. As most curated studies rely on databases for data extraction, the availability of data resources is extremely important. Taking a perspective from pharmacology, extracted data contribute to improved drug treatment outcomes and well-being but with some challenges. Considering available pharmacology literature, it is necessary to review articles and other scientific documents carefully. A typical method of accessing articles on journal websites is through long-established searches. In addition to being labor-intensive, this conventional approach often leads to incomplete-content downloads. This paper presents a new methodology with user-friendly models to accept search keywords according to the investigators' research fields for metadata and full-text articles. To accomplish this, scientifically published records on the pharmacokinetics of drugs were extracted from several sources using our navigating tool called the Web Crawler for Pharmacokinetics (WCPK). The results of metadata extraction provided 74,867 publications for four drug classes. Full-text extractions performed with WCPK revealed that the system is highly competent, extracting over 97% of records. This model helps establish keyword-based article repositories, contributing to comprehensive databases for article curation projects. This paper also explains the procedures adopted to build the proposed customizable-live WCPK, from system design and development to deployment phases.

A Consensus Model of Glucose-Stimulated Insulin Secretion in the Pancreatic β -Cell

The pancreas plays a critical role in maintaining glucose homeostasis through the secretion of hormones from the islets of Langerhans. Glucose-stimulated insulin secretion (GSIS) by the pancreatic β -cell is the main mechanism for reducing elevated plasma glucose. Here we present a systematic modeling workflow for the development of kinetic pathway models using the Systems Biology Markup Language (SBML). Steps include retrieval of information from databases, curation of experimental and clinical data for model calibration and validation, integration of heterogeneous data including absolute and relative measurements, unit normalization, data normalization, and model annotation. An important factor was the reproducibility and exchangeability of the model, which allowed the use of various existing tools. The workflow was applied to construct the first consensus model of GSIS in the pancreatic β -cell based on experimental and clinical data from 39 studies spanning 50 years of pancreatic, islet, and β -cell research in humans, rats, mice, and cell lines. The model consists of detailed glycolysis and equations for insulin secretion coupled to cellular energy state (ATP/ADP ratio). Key findings of our work are that in GSIS there is a glucose-dependent increase in almost all intermediates of glycolysis. This increase in glycolytic metabolites is accompanied by an increase in energy metabolites, especially ATP and NADH. One of the few decreasing metabolites is ADP, which, in combination with the increase in ATP, results in a large increase in ATP/ADP ratios in the β -cell with increasing glucose. Insulin secretion is dependent on ATP/ADP, resulting in glucose-stimulated insulin secretion. The observed glucose-dependent increase in glycolytic intermediates and the resulting change in ATP/ADP ratios and insulin secretion is a robust phenomenon observed across data sets, experimental systems and species. Model predictions of the glucose-dependent response of glycolytic intermediates and insulin secretion are in good agreement with experimental measurements. Our model predicts that factors affecting ATP consumption, ATP formation, hexokinase, phosphofructokinase, and ATP/ADP-dependent insulin secretion have a major effect on GSIS. In conclusion, we have developed and applied a systematic modeling workflow for pathway models that allowed us to gain insight into key mechanisms in GSIS in the pancreatic β -cell.

Machine learning and artificial intelligence in physiologically based pharmacokinetic modeling

Physiologically based pharmacokinetic (PBPK) models are useful tools in drug development and risk assessment of environmental chemicals. PBPK model development requires the collection of species-specific physiological, and chemical-specific absorption, distribution, metabolism, and excretion (ADME) parameters, which can be a time-consuming and expensive process. This raises a need to create computational models capable of predicting input parameter values for PBPK models, especially for new compounds. In this review, we summarize an emerging paradigm for integrating PBPK modeling with machine learning (ML) or artificial intelligence (AI)-based computational methods. This paradigm includes 3 steps (1) obtain time-concentration PK data and/or ADME parameters from publicly available databases, (2) develop ML/AI-based approaches to predict ADME parameters, and (3) incorporate the ML/AI models into PBPK models to predict PK summary statistics (eg, area under the curve and maximum plasma concentration). We also discuss a neural network architecture "neural ordinary differential equation (Neural-ODE)" that is capable of providing better predictive capabilities than other ML methods when used to directly predict time-series PK profiles. In order to support applications of ML/AI methods for PBPK model development, several challenges should be addressed (1) as more data become available, it is important to expand the training set by including the structural diversity of compounds to improve the prediction accuracy of ML/AI models; (2) due to the black box nature of many ML models, lack of sufficient interpretability is a limitation; (3) Neural-ODE has great potential to be used to generate time-series PK profiles for new compounds with limited ADME information, but its application remains to be explored. Despite existing challenges, ML/AI approaches will continue to facilitate the efficient development of robust PBPK models for a large number of chemicals.

Physiologically based pharmacokinetic (PBPK) modeling of the role of CYP2D6 polymorphism for metabolic phenotyping with dextromethorphan

The cytochrome P450 2D6 (CYP2D6) is a key xenobiotic-metabolizing enzyme involved in the clearance of many drugs. Genetic polymorphisms in CYP2D6 contribute to the large inter-individual variability in drug metabolism and could affect metabolic phenotyping of CYP2D6 probe substances such as dextromethorphan (DXM). To study this question, we (i) established an extensive pharmacokinetics dataset for DXM; and (ii) developed and validated a physiologically based pharmacokinetic (PBPK) model of DXM and its metabolites dextrorphan (DXO) and dextrorphan O-glucuronide (DXO-Glu) based on the data. Drug-gene interactions (DGI) were introduced by accounting for changes in CYP2D6 enzyme kinetics depending on activity score (AS), which in combination with AS for individual polymorphisms allowed us to model CYP2D6 gene variants. Variability in CYP3A4 and CYP2D6 activity was modeled based on in vitro data from human liver microsomes. Model predictions are in very good agreement with pharmacokinetics data for CYP2D6 polymorphisms, CYP2D6 activity as described by the AS system, and CYP2D6 metabolic phenotypes (UM, EM, IM, PM). The model was applied to investigate the genotype-phenotype association and the role of CYP2D6 polymorphisms for metabolic phenotyping using the urinary cumulative metabolic ratio (UCMR), DXM/(DXO + DXO-Glu). The effect of parameters on UCMR was studied via sensitivity analysis. Model predictions indicate very good robustness against the intervention protocol (i.e. application form, dosing amount, dissolution rate, and sampling time) and good robustness against physiological variation. The model is capable of estimating the UCMR dispersion within and across populations depending on activity scores. Moreover, the distribution of UCMR and the risk of genotype-phenotype mismatch could be estimated for populations with known CYP2D6 genotype frequencies. The model can be applied for individual prediction of UCMR and metabolic phenotype based on CYP2D6 genotype. Both, model and database are freely available for reuse.

Anti-Inflammatory and Anti-Rheumatic Potential of Selective Plant Compounds by Targeting TLR-4/AP-1 Signaling: A Comprehensive Molecular Docking and Simulation Approaches

Plants are an important source of drug development and numerous plant derived molecules have been used in clinical practice for the ailment of various diseases. The Toll-like receptor-4 (TLR-4) signaling pathway plays a crucial role in inflammation including rheumatoid arthritis. The TLR-4 binds with pro-inflammatory ligands such as lipopolysaccharide (LPS) to induce the downstream signaling mechanism such as nuclear factor κappa B (NF-κB) and mitogen activated protein kinases (MAPKs). This signaling activation leads to the onset of various diseases including inflammation. In the present study, 22 natural compounds were studied against TLR-4/AP-1 signaling, which is implicated in the inflammatory process using a computational approach. These compounds belong to various classes such as methylxanthine, sesquiterpene lactone, alkaloid, flavone glycosides, lignan, phenolic acid, etc. The compounds exhibited different binding affinities with the TLR-4, JNK, NF-κB, and AP-1 protein due to the formation of multiple hydrophilic and hydrophobic interactions. With TLR-4, rutin had the highest binding energy (−10.4 kcal/mol), poncirin had the highest binding energy (−9.4 kcal/mol) with NF-κB and JNK (−9.5 kcal/mol), respectively, and icariin had the highest binding affinity (−9.1 kcal/mol) with the AP-1 protein. The root means square deviation (RMSD), root mean square fraction (RMSF), and radius of gyration (RoG) for 150 ns were calculated using molecular dynamic simulation (MD simulation) based on rutin’s greatest binding energy with TLR-4. The RMSD, RMSF, and RoG were all within acceptable limits in the MD simulation, and the complex remained stable for 150 ns. Furthermore, these compounds were assessed for the potential toxic effect on various organs such as the liver, heart, genotoxicity, and oral maximum toxic dose. Moreover, the blood–brain barrier permeability and intestinal absorption were also predicted using SwissADME software (Lausanne, Switzerland). These compounds exhibited promising physico-chemical as well as drug-likeness properties. Consequently, these selected compounds portray promising anti-inflammatory and drug-likeness properties.

DDPD 1.0: a manually curated and standardized database of digital properties of approved drugs for drug-likeness evaluation and drug development

Drug-likeness is a vital consideration when selecting compounds in the early stage of drug discovery. A series of drug-like properties are needed to predict the drug-likeness of a given compound and provide useful guidelines to increase the likelihood of converting lead compounds into drugs. Experimental physicochemical properties, pharmacokinetic/toxicokinetic properties and maximum dosages of approved small-molecule drugs from multiple text-based unstructured data resources have been manually assembled, curated, further digitized and processed into structured data, which are deposited in the Database of Digital Properties of approved Drugs (DDPD). DDPD 1.0 contains 30 212 drug property entries, including 2250 approved drugs and 32 properties, in a standardized value/unit format. Moreover, two analysis tools are provided to examine the drug-likeness features of given molecules based on the collected property data of approved drugs. Additionally, three case studies are presented to demonstrate how users can utilize the database. We believe that this database will be a valuable resource for the drug discovery and development field.

Database URL: http://www.inbirg.com/ddpd

Pharmacokinetics of Caffeine: A Systematic Analysis of Reported Data for Application in Metabolic Phenotyping and Liver Function Testing

Caffeine is by far the most ubiquitous psychostimulant worldwide found in tea, coffee, cocoa, energy drinks, and many other beverages and food. Caffeine is almost exclusively metabolized in the liver by the cytochrome P-450 enzyme system to the main product paraxanthine and the additional products theobromine and theophylline. Besides its stimulating properties, two important applications of caffeine are metabolic phenotyping of cytochrome P450 1A2 (CYP1A2) and liver function testing. An open challenge in this context is to identify underlying causes of the large inter-individual variability in caffeine pharmacokinetics. Data is urgently needed to understand and quantify confounding factors such as lifestyle (e.g., smoking), the effects of drug-caffeine interactions (e.g., medication metabolized via CYP1A2), and the effect of disease. Here we report the first integrative and systematic analysis of data on caffeine pharmacokinetics from 141 publications and provide a comprehensive high-quality data set on the pharmacokinetics of caffeine, caffeine metabolites, and their metabolic ratios in human adults. The data set is enriched by meta-data on the characteristics of studied patient cohorts and subjects (e.g., age, body weight, smoking status, health status), the applied interventions (e.g., dosing, substance, route of application), measured pharmacokinetic time-courses, and pharmacokinetic parameters (e.g., clearance, half-life, area under the curve). We demonstrate via multiple applications how the data set can be used to solidify existing knowledge and gain new insights relevant for metabolic phenotyping and liver function testing based on caffeine. Specifically, we analyzed 1) the alteration of caffeine pharmacokinetics with smoking and use of oral contraceptives; 2) drug-drug interactions with caffeine as possible confounding factors of caffeine pharmacokinetics or source of adverse effects; 3) alteration of caffeine pharmacokinetics in disease; and 4) the applicability of caffeine as a salivary test substance by comparison of plasma and saliva data. In conclusion, our data set and analyses provide important resources which could enable more accurate caffeine-based metabolic phenotyping and liver function testing.

Prediction of Survival After Partial Hepatectomy Using a Physiologically Based Pharmacokinetic Model of Indocyanine Green Liver Function Tests

The evaluation of hepatic function and functional capacity of the liver are essential tasks in hepatology as well as in hepatobiliary surgery. Indocyanine green (ICG) is a widely applied test compound that is used in clinical routine to evaluate hepatic function. Important questions for the functional evaluation with ICG in the context of hepatectomy are how liver disease such as cirrhosis alters ICG elimination, and if postoperative survival can be predicted from preoperative ICG measurements. Within this work a physiologically based pharmacokinetic (PBPK) model of ICG was developed and applied to the prediction of the effects of a liver resection under various degrees of cirrhosis. For the parametrization of the computational model and validation of model predictions a database of ICG pharmacokinetic data was established. The model was applied (i) to study the effect of liver cirrhosis and liver resection on ICG pharmacokinetics; and (ii) to evaluate the model-based prediction of postoperative ICG-R15 (retention ratio 15 min after administration) as a measure for postoperative outcome. Key results are the accurate prediction of changes in ICG pharmacokinetics caused by liver cirrhosis and postoperative changes of ICG-elimination after liver resection, as validated with a wide range of data sets. Based on the PBPK model, individual survival after liver resection could be classified, demonstrating its potential value as a clinical tool.

Finger sweat analysis enables short interval metabolic biomonitoring in humans

Metabolic biomonitoring in humans is typically based on the sampling of blood, plasma or urine. Although established in the clinical routine, these sampling procedures are often associated with a variety of compliance issues, which are impeding time-course studies. Here, we show that the metabolic profiling of the minute amounts of sweat sampled from fingertips addresses this challenge. Sweat sampling from fingertips is non-invasive, robust and can be accomplished repeatedly by untrained personnel. The sweat matrix represents a rich source for metabolic phenotyping. We confirm the feasibility of short interval sampling of sweat from the fingertips in time-course studies involving the consumption of coffee or the ingestion of a caffeine capsule after a fasting interval, in which we successfully monitor all known caffeine metabolites as well as endogenous metabolic responses. Fluctuations in the rate of sweat production are accounted for by mathematical modelling to reveal individual rates of caffeine uptake, metabolism and clearance. To conclude, metabotyping using sweat from fingertips combined with mathematical network modelling shows promise for broad applications in precision medicine by enabling the assessment of dynamic metabolic patterns, which may overcome the limitations of purely compositional biomarkers.

A decade of machine learning-based predictive models for human pharmacokinetics: Advances and challenges

Traditionally, in vitro and in vivo methods are useful for estimating human pharmacokinetics (PK) parameters; however, it is impractical to perform these complex and expensive experiments on a large number of compounds. The integration of publicly available chemical, or medical Big Data and artificial intelligence (AI)-based approaches led to qualitative and quantitative prediction of human PK of a candidate drug. However, predicting drug response with these approaches is challenging, partially because of the adaptation of algorithmic and limitations related to experimental data. In this report, we provide an overview of machine learning (ML)-based quantitative structure-activity relationship (QSAR) models used in the assessment or prediction of PK values as well as databases available for obtaining such data.

An automated approach to identify scientific publications reporting pharmacokinetic parameters

Pharmacokinetic (PK) predictions of new chemical entities are aided by prior knowledge from other compounds. The development of robust algorithms that improve preclinical and clinical phases of drug development remains constrained by the need to search, curate and standardise PK information across the constantly-growing scientific literature. The lack of centralised, up-to-date and comprehensive repositories of PK data represents a significant limitation in the drug development pipeline.In this work, we propose a machine learning approach to automatically identify and characterise scientific publications reporting PK parameters from in vivo data, providing a centralised repository of PK literature. A dataset of 4,792 PubMed publications was labelled by field experts depending on whether in vivo PK parameters were estimated in the study. Different classification pipelines were compared using a bootstrap approach and the best-performing architecture was used to develop a comprehensive and automatically-updated repository of PK publications. The best-performing architecture encoded documents using unigram features and mean pooling of BioBERT embeddings obtaining an F1 score of 83.8% on the test set. The pipeline retrieved over 121K PubMed publications in which in vivo PK parameters were estimated and it was scheduled to perform weekly updates on newly published articles. All the relevant documents were released through a publicly available web interface (https://app.pkpdai.com) and characterised by the drugs, species and conditions mentioned in the abstract, to facilitate the subsequent search of relevant PK data. This automated, open-access repository can be used to accelerate the search and comparison of PK results, curate ADME datasets, and facilitate subsequent text mining tasks in the PK domain.