HuskinDB, a database for skin permeation of xenobiotics

Artificial intelligence for skin permeability prediction: deep learning

BACKGROUND AND OBJECTIVE

Researchers have put in significant laboratory time and effort in measuring the permeability coefficient (Kp) of xenobiotics. To develop alternative approaches to this labour-intensive procedure, predictive models have been employed by scientists to describe the transport of xenobiotics across the skin. Most quantitative structure-permeability relationship (QSPR) models are derived statistically from experimental data. Recently, artificial intelligence-based computational drug delivery has attracted tremendous interest. Deep learning is an umbrella term for machine-learning algorithms consisting of deep neural networks (DNNs). Distinct network architectures, like convolutional neural networks (CNNs), feedforward neural networks (FNNs), and recurrent neural networks (RNNs), can be employed for prediction.

METHODS

In this project, we used a convolutional neural network, feedforward neural network, and recurrent neural network to predict skin permeability coefficients from a publicly available database reported by Cheruvu et al. The dataset contains 476 records of 145 chemicals, xenobiotics, and pharmaceuticals, administered on the human epidermis in vitro from aqueous solutions of constant concentration either saturated in infinite dose quantities or diluted. All the computations were conducted with Python under Anaconda and Jupyterlab environment after importing the required Python, Keras, and Tensorflow modules.

RESULTS

We used a convolutional neural network, feedforward neural network, and recurrent neural network to predict log kp.

CONCLUSION

This research work shows that deep learning networks can be successfully used to digitally screen and predict the skin permeability of xenobiotics.

Exploring the Therapeutic Potential of Petiveria alliacea L. Phytochemicals: A Computational Study on Inhibiting SARS-CoV-2's Main Protease (Mpro)

The outbreak of SARS-CoV-2, also known as the COVID-19 pandemic, is still a critical risk factor for both human life and the global economy. Although, several promising therapies have been introduced in the literature to inhibit SARS-CoV-2, most of them are synthetic drugs that may have some adverse effects on the human body. Therefore, the main objective of this study was to carry out an in-silico investigation into the medicinal properties of Petiveria alliacea L. (P. alliacea L.)-mediated phytocompounds for the treatment of SARS-CoV-2 infections since phytochemicals have fewer adverse effects compared to synthetic drugs. To explore potential phytocompounds from P. alliacea L. as candidate drug molecules, we selected the infection-causing main protease (Mpro) of SARS-CoV-2 as the receptor protein. The molecular docking analysis of these receptor proteins with the different phytocompounds of P. alliacea L. was performed using AutoDock Vina. Then, we selected the three top-ranked phytocompounds (myricitrin, engeletin, and astilbin) as the candidate drug molecules based on their highest binding affinity scores of -8.9, -8.7 and -8.3 (Kcal/mol), respectively. Then, a 100 ns molecular dynamics (MD) simulation study was performed for their complexes with Mpro using YASARA software, computed RMSD, RMSF, PCA, DCCM, MM/PBSA, and free energy landscape (FEL), and found their almost stable binding performance. In addition, biological activity, ADME/T, DFT, and drug-likeness analyses exhibited the suitable pharmacokinetics properties of the selected phytocompounds. Therefore, the results of this study might be a useful resource for formulating a safe treatment plan for SARS-CoV-2 infections after experimental validation in wet-lab and clinical trials.

An update of skin permeability data based on a systematic review of recent research

The cutaneous absorption parameters of xenobiotics are crucial for the development of drugs and cosmetics, as well as for assessing environmental and occupational chemical risks. Despite the great variability in the design of experimental conditions due to uncertain international guidelines, datasets like HuskinDB have been created to report skin absorption endpoints. This review updates available skin permeability data by rigorously compiling research published between 2012 and 2021. Inclusion and exclusion criteria have been selected to build the most harmonized and reusable dataset possible. The Generative Topographic Mapping method was applied to the present dataset and compared to HuskinDB to monitor the progress in skin permeability research and locate chemotypes of particular concern. The open-source dataset (SkinPiX) includes steady-state flux, maximum flux, lag time and permeability coefficient results for the substances tested, as well as relevant information on experimental parameters that can impact the data. It can be used to extract subsets of data for comparisons and to build predictive models.

Methods for the characterisation of dermal uptake: Progress and perspectives for organophosphate esters

Organophosphate esters (OPEs) are a group of pollutants that are widely detected in the environment at high concentrations. They can adversely affect human health through multiple routes of exposure, including dermal uptake. Although attention has been paid to achieving an accurate and complete quantification of the dermal uptake of OPEs, existing evaluation methods and parameters have obvious weaknesses. This study reviewed two main categories of methodologies, namely the relative absorption (RA) model and the permeability coefficient (PC) model, which are widely used to assess the dermal uptake of OPEs. Although the PC model is more accurate and is increasingly used, the most important parameter in this model, the permeability coefficient (Kp), has been poorly characterised for OPEs, resulting in considerable errors in the estimation of the dermal uptake of OPEs. Thus, the detailed in vitro methods for the determination of Kp are summarised and sorted. Furthermore, the commonly used skin membranes are identified and the factors affecting Kp and corresponding mechanisms are discussed. In addition, the experimental conditions, conclusions, and available data on Kp values of the OPEs are thoroughly summarised. Finally, the corresponding knowledge gaps are proposed, and a more accurate and sophisticated experimental system and unknown Kp values for OPEs are suggested.

An examination of published datasets of skin permeability and partition coefficients

Permeability and partition coefficients of the skin barrier are important for assessing dermal absorption, bioavailability, and safety of cosmetics and medicine. We use the Potts and Guy equation to analyse the dependence of skin permeability on the hydrophobicity of permeants and highlight the significant differences in published datasets. Correlations of solute partition to skin are examined to understand the likely causes of the differences in the skin permeability datasets. Recently published permeability datasets show weak correlation and low dependence on hydrophobicity. As expected, early datasets show good correlation with hydrophobicity due to the related derivation. The weaker correlation of later datasets cannot be explained by the partition to skin lipids. All the datasets of solute partition to skin lipid showed a similar correlation to hydrophobicity where the log-linear correlation coefficient of partition is almost the same of the log-linear coefficient of Potts and Guy equation. Weak correlation and dependence of the late permeability datasets with SC lipid/water partition and that they are significantly under predicted by the Potts and Guy equation suggests either additional non-lipid pathway at play or a weaker skin barrier property.

New approaches build upon historical studies in dermal toxicology

These are my personal reflections on the history of approaches to understanding dermal toxicology brought together for the Paton Prize Award. This is not a comprehensive account of all publications from in vivo studies in humans to development of in vitro and in silico approaches but highlghts important progress. I will consider what is needed now to influence approaches to understanding dermal exposure with the current development and use of NAMs (new approach methodologies).

Fragment contribution models for predicting skin permeability using HuskinDB

Mathematical models to predict skin permeation tend to be based on animal derived experimental data as well as knowing physicochemical properties of the compound under investigation, such as molecular volume, polarity and lipophilicity. This paper presents a strikingly contrasting model to predict permeability, formed entirely from simple chemical fragment (functional group) data and a recently released, freely accessible human (i.e. non-animal) skin permeation database, known as the 'Human Skin Database - HuskinDB'. Data from within the database allowed development of several fragment-based models, each including a calculable effect for all of the most commonly encountered functional groups present in compounds within the database. The developed models can be applied to predict human skin permeability (logKp) for any compound containing one or more of the functional groups analysed from the dataset with no need to know any other physicochemical properties, solely the type and number of each functional group within the chemical structure itself. This approach simplifies mathematical prediction of permeability for compounds with similar properties to those used in this study.

Potential for Machine Learning to Address Data Gaps in Human Toxicity and Ecotoxicity Characterization

Machine Learning (ML) is increasingly applied to fill data gaps in assessments to quantify impacts associated with chemical emissions and chemicals in products. However, the systematic application of ML-based approaches to fill chemical data gaps is still limited, and their potential for addressing a wide range of chemicals is unknown. We prioritized chemical-related parameters for chemical toxicity characterization to inform ML model development based on two criteria: (1) each parameter's relevance to robustly characterize chemical toxicity described by the uncertainty in characterization results attributable to each parameter and (2) the potential for ML-based approaches to predict parameter values for a wide range of chemicals described by the availability of chemicals with measured parameter data. We prioritized 13 out of 38 parameters for developing ML-based approaches, while flagging another nine with critical data gaps. For all prioritized parameters, we performed a chemical space analysis to assess further the potential for ML-based approaches to predict data for diverse chemicals considering the structural diversity of available measured data, showing that ML-based approaches can potentially predict 8-46% of marketed chemicals based on 1-10% with available measured data. Our results can systematically inform future ML model development efforts to address data gaps in chemical toxicity characterization.

Immobilized Keratin HPLC Stationary Phase-A Forgotten Model of Transdermal Absorption: To What Molecular and Biological Properties Is It Relevant?

Chromatographic retention data collected on immobilized keratin (KER) or immobilized artificial membrane (IAM) stationary phases were used to predict skin permeability coefficient (log K_p) and bioconcentration factor (log BCF) of structurally unrelated compounds. Models of both properties contained, apart from chromatographic descriptors, calculated physico-chemical parameters. The log K_p model, containing keratin-based retention factor, has slightly better statistical parameters and is in a better agreement with experimental log K_p data than the model derived from IAM chromatography; both models are applicable primarily to non-ionized compounds.Based on the multiple linear regression (MLR) analyses conducted in this study, it was concluded that immobilized keratin chromatographic support is a moderately useful tool for skin permeability assessment.However, chromatography on immobilized keratin may also be of use for a different purpose-in studies of compounds' bioconcentration in aquatic organisms.

Nutricosmetics: A new frontier in bioactive peptides' research toward skin aging

Food derived bioactive peptides are small protein fragments (2-20 amino acids long) that can exhibit health benefits, beyond basic nutrition. For example, food bioactive peptides can act as physiological modulators with hormone or drug-like activities including anti-inflammatory, antimicrobial, antioxidant, and the ability to inhibit enzymes related to chronic disease metabolism. Recently, bioactive peptides have been studied for their potential role as nutricosmetics. For example, bioactive peptides can impart skin-aging protection toward extrinsic (i.e., environmental and sun UV-ray damage) and intrinsic (i.e., natural cell or chronological aging) factors. Specifically, bioactive peptides have demonstrated antioxidant and antimicrobial activates toward reactive oxygen species (ROS) and pathogenic bacteria associated with skin diseases, respectively. The anti-inflammatory properties of bioactive peptides using in vivo models has also been reported, where peptides have shown to decreased the expression of IL-6, TNF-α, IL-1β, interferon-γ (INF-γ), and interleukin-17 (IL-17) in mice models. This chapter will discuss the main factors that trigger skin-aging processes, as well as provide examples of in vitro, in vivo, and in silico applications of bioactive peptides in relation to nutricosmetic applications.

Predicting skin permeability using HuskinDB

A freely accessible database has recently been released that provides measurements available in the literature on human skin permeation data, known as the ‘Human Skin Database – HuskinDB’. Although this database is extremely useful for sourcing permeation data to help with toxicity and efficacy determination, it cannot be beneficial when wishing to consider unlisted, or novel compounds. This study undertakes analysis of the data from within HuskinDB to create a model that predicts permeation for any compound (within the range of properties used to create the model). Using permeability coefficient (K_p) data from within this resource, several models were established for K_p values for compounds of interest by varying the experimental parameters chosen and using standard physicochemical data. Multiple regression analysis facilitated creation of one particularly successful model to predict K_p through human skin based only on three chemical properties. The model transforms the dataset from simply a resource of information to a more beneficial model that can be used to replace permeation testing for a wide range of compounds.

A curated binary pattern multitarget dataset of focused ATP-binding cassette transporter inhibitors

Multitarget datasets that correlate bioactivity landscapes of small-molecules toward different related or unrelated pharmacological targets are crucial for novel drug design and discovery. ATP-binding cassette (ABC) transporters are critical membrane-bound transport proteins that impact drug and metabolite distribution in human disease as well as disease diagnosis and therapy. Molecular-structural patterns are of the highest importance for the drug discovery process as demonstrated by the novel drug discovery tool ‘computer-aided pattern analysis’ (‘C@PA’). Here, we report a multitarget dataset of 1,167 ABC transporter inhibitors analyzed for 604 molecular substructures in a statistical binary pattern distribution scheme. This binary pattern multitarget dataset (ABC_BPMDS) can be utilized for various areas. These areas include the intended design of (i) polypharmacological agents, (ii) highly potent and selective ABC transporter-targeting agents, but also (iii) agents that avoid clearance by the focused ABC transporters [e.g., at the blood-brain barrier (BBB)]. The information provided will not only facilitate novel drug prediction and discovery of ABC transporter-targeting agents, but also drug design in general in terms of pharmacokinetics and pharmacodynamics.

Measurement(s)	Influx • Efflux • Tracer • Transport velocity
Technology Type(s)	Fluorometry • Radioactivity • Plate reader • Flow cytometer • Tracer distribution
Factor Type(s)	half-maximal inhibition concentration
Sample Characteristic - Organism	Homo sapiens
Sample Characteristic - Environment	cell culture
Sample Characteristic - Location	Kingdom of Norway • Germany • Australia • Latvia

Studying and mitigating the effects of data drifts on ML model performance at the example of chemical toxicity data

Machine learning models are widely applied to predict molecular properties or the biological activity of small molecules on a specific protein. Models can be integrated in a conformal prediction (CP) framework which adds a calibration step to estimate the confidence of the predictions. CP models present the advantage of ensuring a predefined error rate under the assumption that test and calibration set are exchangeable. In cases where the test data have drifted away from the descriptor space of the training data, or where assay setups have changed, this assumption might not be fulfilled and the models are not guaranteed to be valid. In this study, the performance of internally valid CP models when applied to either newer time-split data or to external data was evaluated. In detail, temporal data drifts were analysed based on twelve datasets from the ChEMBL database. In addition, discrepancies between models trained on publicly-available data and applied to proprietary data for the liver toxicity and MNT in vivo endpoints were investigated. In most cases, a drastic decrease in the validity of the models was observed when applied to the time-split or external (holdout) test sets. To overcome the decrease in model validity, a strategy for updating the calibration set with data more similar to the holdout set was investigated. Updating the calibration set generally improved the validity, restoring it completely to its expected value in many cases. The restored validity is the first requisite for applying the CP models with confidence. However, the increased validity comes at the cost of a decrease in model efficiency, as more predictions are identified as inconclusive. This study presents a strategy to recalibrate CP models to mitigate the effects of data drifts. Updating the calibration sets without having to retrain the model has proven to be a useful approach to restore the validity of most models.

Physicochemical properties determining drug detection in skin

Chemicals, including some systemically administered xenobiotics and their biotransformations, can be detected noninvasively using skin swabs and untargeted metabolomics analysis. We sought to understand the principal drivers that determine whether a drug taken orally or systemically is likely to be observed on the epidermis by using a random forest classifier to predict which drugs would be detected on the skin. A variety of molecular descriptors describing calculated properties of drugs, such as measures of volume, electronegativity, bond energy, and electrotopology, were used to train the classifier. The mean area under the receiver operating characteristic curve was 0.71 for predicting drug detection on the epidermis, and the SHapley Additive exPlanations (SHAP) model interpretation technique was used to determine the most relevant molecular descriptors. Based on the analysis of 2561 US Food and Drug Administration (FDA)‐approved drugs, we predict that therapeutic drug classes, such as nervous system drugs, are more likely to be detected on the skin. Detecting drugs and other chemicals noninvasively on the skin using untargeted metabolomics could be a useful clinical advancement in therapeutic drug monitoring, adherence, and health status.

Reassessment of the experimental skin permeability coefficients of polycyclic aromatic hydrocarbons and organophosphorus pesticides

Human exposure to polycyclic aromatic hydrocarbons (PAHs) and organophosphorus pesticides (OPPs) by dermal route is a continuing concern in environmental and occupational toxicology. Diverse authors have measured in vitro the absorption flux and permeability coefficient (k_P) of those compounds delivered on skin surface using volatile solvents. However, there isn't a harmonized method to obtain k_P when the test substance is deposited on the skin as a solid. Consequently, varied experimental k_Ps have been reported for PAHs and OPPs, most in clear disagreement with the values predicted by well-established mathematical models. In this work, we collected the permeation fluxes reported for these toxicants through human skin and calculated the (aqueous) k_Ps using a method based on the maximum flux and water solubility. The reanalyzed fluxes and recalculated k_Ps show improved consistency between the different experimental works and mathematical models. Notably, the recalculated k_P of benzo[a]pyrene, among others, was approximately 100 times higher than it had been previously considered. Suggestions are given to generalize the method in studies with other solvent-deposited toxicants and drugs.