Scholarlink

<p>The development of pharmaceutical formulations typically adopts a lengthy and costly trial-and-error approach, often yielding inaccurate predictions of effectiveness and safety of drug-delivery systems, including hydrogels for antibiotics. Accordingly, machine learning (ML) has emerged as a useful method for predictions based on experimental data. ML can predict a numerical value through numerous supervised models, which are trained and assessed to determine the optimal option. Upon attaining the desired accuracy, the selected model can be applied for prospective predictions and interpreted to extract useful insights. The aim of our study was to apply a hybrid ML approach to predict the release profiles of an antibiotic (silver sulfadiazine) from temperature-responsive hydrogels based on in vitro data. The study explored hydrogel formulations of varying PF-127 and cellulose percentages, temperatures, and drug concentrations. Under this hybrid approach, ML models were investigated alongside different kinetics and mechanisms models. Six ML models—random forest, Gaussian regressor, linear regression, MLP regressor, support vector machine, and kernel ridge—were adopted to predict experimental drug-release data. Model performances were evaluated through the correlation coefficient (R<sup>2</sup>) and mean absolute percentage error (MAPE). We found that the random forest model exhibited a superior performance, achieving an R<sup>2</sup> of 0.99 and MAPE 0.002, indicating a robust fit to the data. The release half-life (t<sub>50%</sub>) increased as temperature rose from 18 to 32&#xa0;°C, then decreased at 40&#xa0;°C, while increasing the drug percentage and polymer concentration prolonged t<sub>50%</sub>. Zero-order and Higuchi kinetic models best fit the data, with non-Fickian diffusion and Super Case II mechanisms dominating. These findings demonstrate the potential of ML to streamline pharmaceutical development, reducing the need for extensive laboratory trials.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A hybrid machine learning framework for predicting drug-release profiles, kinetics, and mechanisms of temperature-responsive hydrogels

  • 期刊论文

  • 发表日期：

    2025年1月13日

查看全文
  • Maha Mohammad AL-Rajabi,
  • Samer Alzyod,
  • Akshay Patel,
  • Yeit Haan Teow

摘要

The development of pharmaceutical formulations typically adopts a lengthy and costly trial-and-error approach, often yielding inaccurate predictions of effectiveness and safety of drug-delivery systems, including hydrogels for antibiotics. Accordingly, machine learning (ML) has emerged as a useful method for predictions based on experimental data. ML can predict a numerical value through numerous supervised models, which are trained and assessed to determine the optimal option. Upon attaining the desired accuracy, the selected model can be applied for prospective predictions and interpreted to extract useful insights. The aim of our study was to apply a hybrid ML approach to predict the release profiles of an antibiotic (silver sulfadiazine) from temperature-responsive hydrogels based on in vitro data. The study explored hydrogel formulations of varying PF-127 and cellulose percentages, temperatures, and drug concentrations. Under this hybrid approach, ML models were investigated alongside different kinetics and mechanisms models. Six ML models—random forest, Gaussian regressor, linear regression, MLP regressor, support vector machine, and kernel ridge—were adopted to predict experimental drug-release data. Model performances were evaluated through the correlation coefficient (R2) and mean absolute percentage error (MAPE). We found that the random forest model exhibited a superior performance, achieving an R2 of 0.99 and MAPE 0.002, indicating a robust fit to the data. The release half-life (t50%) increased as temperature rose from 18 to 32 °C, then decreased at 40 °C, while increasing the drug percentage and polymer concentration prolonged t50%. Zero-order and Higuchi kinetic models best fit the data, with non-Fickian diffusion and Super Case II mechanisms dominating. These findings demonstrate the potential of ML to streamline pharmaceutical development, reducing the need for extensive laboratory trials.