Machine learning methods for prediction of food effects on bioavailability: A comparison of support vector machines and artificial neural networks

Enhancing plant-based cheese formulation through molecular docking and dynamic simulation of tocopherol and retinol complexes with zein, soy and almond proteins via SVM-machine learning integration

The current study addresses the growing demand for sustainable plant-based cheese alternatives by employing molecular docking and deep learning algorithms to optimize protein-ligand interactions. Focusing on key proteins (zein, soy, and almond protein) along with tocopherol and retinol, the goal was to improve texture, nutritional value, and flavor characteristics via dynamic simulations. The findings demonstrated that the docking analysis presented high accuracy in predicting conformational changes. Flexible docking algorithms provided insights into dynamic interactions, while analysis of energetics revealed variations in binding strengths. Tocopherol exhibited stronger affinity (-5.8Kcal/mol) to zein compared to retinol (-4.1Kcal/mol). Molecular dynamics simulations offered comprehensive insights into stability and behavior over time. The integration of machine learning algorithms improved the classification and the prediction accuracy, achieving a rate of 71.59%. This study underscores the significance of molecular understanding in driving innovation in the plant-based cheese industry, facilitating the development of sustainable alternatives to traditional dairy products.

DDID: a comprehensive resource for visualization and analysis of diet-drug interactions

Diet-drug interactions (DDIs) are pivotal in drug discovery and pharmacovigilance. DDIs can modify the systemic bioavailability/pharmacokinetics of drugs, posing a threat to public health and patient safety. Therefore, it is crucial to establish a platform to reveal the correlation between diets and drugs. Accordingly, we have established a publicly accessible online platform, known as Diet-Drug Interactions Database (DDID, https://bddg.hznu.edu.cn/ddid/), to systematically detail the correlation and corresponding mechanisms of DDIs. The platform comprises 1338 foods/herbs, encompassing flora and fauna, alongside 1516 widely used drugs and 23 950 interaction records. All interactions are meticulously scrutinized and segmented into five categories, thereby resulting in evaluations (positive, negative, no effect, harmful and possible). Besides, cross-linkages between foods/herbs, drugs and other databases are furnished. In conclusion, DDID is a useful resource for comprehending the correlation between foods, herbs and drugs and holds a promise to enhance drug utilization and research on drug combinations.

Advancing algorithmic drug product development: Recommendations for machine learning approaches in drug formulation

Artificial intelligence is a rapidly expanding area of research, with the disruptive potential to transform traditional approaches in the pharmaceutical industry, from drug discovery and development to clinical practice. Machine learning, a subfield of artificial intelligence, has fundamentally transformed in silico modelling and has the capacity to streamline clinical translation. This paper reviews data-driven modelling methodologies with a focus on drug formulation development. Despite recent advances, there is limited modelling guidance specific to drug product development and a trend towards suboptimal modelling practices, resulting in models that may not give reliable predictions in practice. There is an overwhelming focus on benchtop experimental outcomes obtained for a specific modelling aim, leaving the capabilities of data scraping or the use of combined modelling approaches yet to be fully explored. Moreover, the preference for high accuracy can lead to a reliance on black box methods over interpretable models. This further limits the widespread adoption of machine learning as black boxes yield models that cannot be easily understood for the purposes of enhancing product performance. In this review, recommendations for conducting machine learning research for drug product development to ensure trustworthiness, transparency, and reliability of the models produced are presented. Finally, possible future directions on how research in this area might develop are discussed to aim for models that provide useful and robust guidance to formulators.

Analysis of the interplay of physiological response to food intake and drug properties in food-drug interactions

The effect of food on oral drug absorption is determined by the complex interplay among gut physiological factors and drug properties. The currently used dissolution testing and classification systems (biopharmaceutics classification system, BCS or biopharmaceutics drug disposition classification system, BDDCS) do not account for dynamic changes in gastrointestinal physiology caused by food intake. This study aimed to identify key drug properties that influence food effect (FE) using supervised machine learning approaches. The analysis showed that drugs with high logP, dose number, and extraction ratio have a higher probability of positive FE, while drugs with low permeability and high efflux saturation index have a greater likelihood of negative FE. Weakly acidic drugs also showed a greater probability of positive FE, particularly at pKa >4.3. The importance of drug properties in predicting FE was ranked as logP, dose number, extraction ratio, pKa, and permeability. The accuracy of FE prediction using the models was compared with BCS and extended clearance classification system (ECCS). Overall, the likelihood or magnitude of FE depends on physiological changes to food intake such as altered bile acid secretion rate, intestinal metabolism, transport kinetics, and gastric emptying time, which should be considered along with drug properties (e.g., solubility, logP, and ionization) in predicting FE of orally administered drugs.

Machine learning-driven QSAR models for predicting the mixture toxicity of nanoparticles

Research on theoretical prediction methods for the mixture toxicity of engineered nanoparticles (ENPs) faces significant challenges. The application of in silico methods based on machine learning is emerging as an effective strategy to address the toxicity prediction of chemical mixtures. Herein, we combined toxicity data generated in our lab with experimental data reported in the literature to predict the combined toxicity of seven metallic ENPs for Escherichia coli at different mixing ratios (22 binary combinations). We thereafter applied two machine learning (ML) techniques, support vector machine (SVM) and neural network (NN), and compared the differences in the ability to predict the combined toxicity by means of the ML-based methods and two component-based mixture models: independent action and concentration addition. Among 72 developed quantitative structure-activity relationship (QSAR) models by the ML methods, two SVM-QSAR models and two NN-QSAR models showed good performance. Moreover, an NN-based QSAR model combined with two molecular descriptors, namely enthalpy of formation of a gaseous cation and metal oxide standard molar enthalpy of formation, showed the best predictive power for the internal dataset (R²_test = 0.911, adjusted R²_test = 0.733, RMSE_test = 0.091, and MAE_test = 0.067) and for the combination of internal and external datasets (R²_test = 0.908, adjusted R²_test = 0.871, RMSE_test = 0.255, and MAE_test = 0.181). In addition, the developed QSAR models performed better than the component-based models. The estimation of the applicability domain of the selected QSAR models showed that all the binary mixtures in training and test sets were in the applicability domain. This study approach could provide a methodological and theoretical basis for the ecological risk assessment of mixtures of ENPs.

Assessment of food effects during clinical development

Food-drug interactions frequently hamper oral drug development due to various physicochemical, physiological and formulation-dependent mechanisms. This has stimulated the development of a range of promising biopharmaceutical assessment tools which, however, lack standardized settings and protocols. Hence, this manuscript aims to provide an overview of the general approach and the methodology used in food effect assessment and prediction. For in vitro dissolution-based predictions, the expected food effect mechanism should be carefully considered when selecting the level of complexity of the model, together with its drawbacks and advantages. Typically, in vitro dissolution profiles are then incorporated into physiologically based pharmacokinetic models, which can estimate the impact of food-drug interactions on bioavailability within 2-fold prediction error, at least. Positive food effects related to drug solubilization in the GI tract are easier to predict than negative food effects. Preclinical animal models also provide a good level of food effect prediction, with beagle dogs remaining the gold standard. When solubility-related food-drug interactions have large clinical impact, advanced formulation approaches can be used to improve fasted state pharmacokinetics, hence decreasing the fasted/fed difference in oral bioavailability. Finally, the knowledge from all studies should be combined to secure regulatory approval of the labelling instructions.

Selected Aspects of the Analytical and Pharmaceutical Profiles of Nifurtimox

Nifurtimox is a nitroheterocyclic drug employed for treatment of trypanosomiases (Chagas disease and West African sleeping sickness); its use for certain cancers has also been assessed. Despite having been in the market for over 50 years, knowledge of nifurtimox is still fragmentary and incomplete. Relevant aspects of the chemistry and biology of nifurtimox are reviewed to summarize the current knowledge of this drug. These comprise its chemical synthesis and the preparation of some analogues, as well as its chemical degradation. Selected physical data and physicochemical properties are also listed, along with different approaches toward the analytical characterization of the drug, including electrochemical (polarography, cyclic voltammetry), spectroscopic (ultraviolet-visible, nuclear magnetic resonance, electron spin resonance), and single crystal X-ray diffractometry. The array of polarographic, ultraviolet-visible spectroscopic, and chromatographic methods available for the analytical determination of nifurtimox (in bulk drug, pharmaceutical formulations, and biological samples), are also presented and discussed, along with chiral chromatographic and electrophoretic alternatives for the separation of the enantiomers of the drug. Aspects of the drug likeliness of nifurtimox, its classification in the Biopharmaceutical Classification System, and available pharmaceutical formulations are detailed, whereas pharmacological, chemical, and biological aspects of its metabolism and disposition are discussed.

Comparison of a single pharmaceutical surfactant versus intestinal biorelevant media for etravirine dissolution: Role and impact of micelle diffusivity

Etravirine is an antiviral whose oral absorption is limited by low solubility/dissolution. The objective was to predict and compare etravirine's surfactant-mediated dissolution into polyoxyethylene-10 lauryl ether (POE) and FeSSIF-V2, including the contribution of slow micelle diffusivity. Dynamic light scattering (DLS) was used to measure the size and diffusivity values of drug-loaded micelles. In vitro intrinsic dissolution into surfactant media were predicted using a model for surfactant-mediated dissolution. Compared to maleic buffer, POE and FeSSIF-V2 increased etravirine solubility 232-fold and 8.97-fold, respectively. From DLS, micelle diffusivity of drug-loaded POE micelle and FeSSIF-V2 mixed-micelle was 5.15x10^-7 cm²/s and 5.76x10^-8 cm²/s, respectively. Observed and predicted dissolution enhancement into POE were 50.7 and 31.3, and 1.26 and 1.24 into FeSSIF-V2, respectively. Hence, there was high dissolution enhancement into POE, although the observed enhancement was only 21.9% of the observed solubility enhancement, reflecting the attenuating impact of the large and slowly diffusing drug-loaded POE micelles. Meanwhile, there was minimal dissolution enhancement into FeSSIF-V2, and the observed enhancement was only 14.0% of the observed solubility enhancement, reflecting the even slower diffusing drug-loaded FeSSIF-V2 mixed-micelles compared to drug-loaded POE micelles. Results are considered in light of designing a single pharmaceutical surfactant system for dissolution that mimics a FeSSIF-V2 system.

Predicting Bile and Lipid Interaction for Drug Substances

Predicting biopharmaceutical characteristics and food effects for drug substances may substantially leverage rational formulation outcomes. We established a bile and lipid interaction prediction model for new drug substances and further explored the model for the prediction of bile-related food effects. One hundred and forty-one drugs were categorized as bile and/or lipid interacting and noninteracting drugs using ¹H nuclear magnetic resonance (NMR) spectroscopy. Quantitative structure-property relationship modeling with molecular descriptors was applied to predict a drug's interaction with bile and/or lipids. Bile interaction, for example, was indicated by two descriptors characterizing polarity and lipophilicity with a high balanced accuracy of 0.8. Furthermore, the predicted bile interaction correlated with a positive food effect. Reliable prediction of drug substance interaction with lipids required four molecular descriptors with a balanced accuracy of 0.7. These described a drug's shape, lipophilicity, aromaticity, and hydrogen bond acceptor capability. In conclusion, reliable models might be found through drug libraries characterized for bile interaction by NMR. Furthermore, there is potential for predicting bile-related positive food effects.

Bio-enabling strategies to mitigate the pharmaceutical food effect: a mini review

The concomitant administration of oral drugs with food can result in significant changes in bioavailability, leading to variable pharmacokinetics and considerable clinical implications, such as over- or under-dosing. Consequently, there is increasing demand for bio-enabling formulation strategies to reduce variability in exposure between the fasted and fed state and/or mitigate the pharmaceutical food effect. The current review critically evaluates technologies that have been implemented to overcome the positive food effects of pharmaceutical drugs, including, lipid-based formulations, nanosized drug preparations, cyclodextrins, amorphisation and solid dispersions, prodrugs and salts. Additionally, improved insight into preclinical models for predicting the food effect is provided. Despite the wealth of research, this review demonstrates that application of optimal formulation strategies to mitigate the positive food effects and the evaluation in preclinical models is not a universal approach, and improved standardisation of models to predict the food effects would be desirable. Ultimately, the successful reformulation of specific drugs to eliminate the food effect provides a panoply of advantages for patients with regard to clinical efficacy and compliance.

Development of In Vitro Dissolution Testing Methods to Simulate Fed Conditions for Immediate Release Solid Oral Dosage Forms

In vitro dissolution testing is widely used to mimic and predict in vivo performance of oral drug products in the gastrointestinal (GI) tract. This literature review assesses the current in vitro dissolution methodologies being employed to simulate and predict in vivo drug dissolution under fasted and fed conditions, with emphasis on immediate release (IR) solid oral dosage forms. Notable human GI physiological conditions under fasted and fed states have been reviewed and summarized. Literature results showed that dissolution media, mechanical forces, and transit times are key dissolution test parameters for simulating specific postprandial conditions. A number of biorelevant systems, including the fed stomach model (FSM), GastroDuo device, dynamic gastric model (DGM), simulated gastrointestinal tract models (TIM), and the human gastric simulator (HGS), have been developed to mimic the postprandial state of the stomach. While these models have assisted in expanding physiological relevance of in vitro dissolution tests, in general, these models lack the ability to fully replicate physiological conditions/processes. Furthermore, the translatability of in vitro data to an in vivo system remains challenging. Additionally, physiologically based pharmacokinetic (PBPK) modeling has been employed to evaluate the effect of food on drug bioavailability and bioequivalence. Here, we assess the current status of in vitro dissolution methodologies and absorption PBPK modeling approaches to identify knowledge gaps and facilitate further development of in vitro dissolution methods that factor in fasted and fed states. Prediction of in vivo drug performance under fasted and fed conditions via in vitro dissolution testing and modeling may potentially help efforts in harmonizing global regulatory recommendations regarding in vivo fasted and fed bioequivalence studies for solid oral IR products.

Machine learning predicts the effect of food on orally administered medicines

Food-mediated changes to drug absorption, termed the food effect, are hard to predict and can have significant implications for the safety and efficacy of oral drug products in patients. Mimicking the prandial states of the human gastrointestinal tract in preclinical studies is challenging, poorly predictive and can produce difficult to interpret datasets. Machine learning (ML) has emerged from the computer science field and shows promise in interpreting complex datasets present in the pharmaceutical field. A ML-based approach aimed to predict the food effect based on an extensive dataset of over 311 drugs with more than 20 drug physicochemical properties, referred to as features. Machine learning techniques were tested; including logistic regression, support vector machine, k-Nearest neighbours and random forest. First a standard ML pipeline using a 80:20 split for training and testing was tried to predict no food effect, negative food effect and positive food effect, however this lead to specificities of less than 40%. To overcome this, a strategic ML pipeline was devised and three tasks were developed. Random forest achieved the strongest performance overall. High accuracies and sensitivities of 70%, 80% and 70% and specificities of 71%, 76% and 71% were achieved for classifying; (i) no food effect vs food effect, (ii) negative food vs positive food effect and (iii) no food effect vs negative food effect vs positive food effect, respectively. Feature importance using random forest ranked the features by importance for building the predictive tasks. The calculated dose number was the most important feature. Here, ML has provided an effective screening tool for predicting the food effect, with the potential to select lead compounds with no food effect, reduce the number of animal studies, and accelerate oral drug development studies.