Scholarlink

Purpose <p>The aim of this project was to predict the skin permeability of compounds in three modified datasets (Steinmetz et al., Stevens et al. as well as Wilschut et al.).</p> Methods <p>We employed Elasticnet, Ridge and Decision Tree Regression algorithms to forecast the skin permeability values of these compounds.</p> Results <p>When the Ridge regression technique was applied to the modified Wilschut et al dataset, the mean squared error was 0.20, the mean absolute error was 0.33 and the coefficient of determination (R²) was 0.61. The application of the same technique to the modified Stevens et al. dataset resulted in a mean squared error of 1.04, a mean absolute error of 0.5172 and an R-squared value of 0.18. The utilization of the Ridge regression method on the modified Steinmetz et al. dataset resulted in a mean squared error of 0.65, a mean absolute error of 0.67 and an R-squared Score of 0.48. When the Elasticnet regression approach was used on the modified Wilschut et al, the mean squared error was 0.24 and the coefficient of determination (R²) was 0.60. The utilization of the Elasticnet regression technique on the modified Steinmetz et al dataset led to a mean squared error of 0.88 and the coefficient of determination (R²) of 0.30. In comparison, Elasticnet regression technique on the modified Stevens et al dataset led to a mean squared error of 0.32 and the coefficient of determination (R²) of 0.42. The utilization of the Decision Tree(DT) regression on the modified Wilschut et al. dataset, resulted in the mean squared error of 0.28 and the coefficient of determination (R²) of 0.53. Decision Tree regression technique on the modified Stevens et al. dataset yielded a mean squared error: 0.31 and an R-squared Score :of 0.66. When the DT regression method was used on the Steinmetz et al dataset, the mean squared error was 0.89 and the R-squared Score was 0.14.</p> Conclusion <p>Our comparison analysis utilizing ElasticNet, Ridge, and Decision Tree regression models to forecast skin permeability across three datasets provides significant insights into the relationship between data quality and model efficacy. This finding is consistent with and enhances the advancing field of computer modeling in cutaneous absorption, specifically in medication development, cosmetic safety, and regulatory science.</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

Predicting Skin Permeability of Compounds with Elasticnet, Ridge and Decision Tree Regression Methods

  • 期刊论文

  • 发表日期：

    2025年6月15日

查看全文
  • Kevin Ita,
  • Pegah Capaul,
  • Pardis Khani

摘要

Purpose

The aim of this project was to predict the skin permeability of compounds in three modified datasets (Steinmetz et al., Stevens et al. as well as Wilschut et al.).

Methods

We employed Elasticnet, Ridge and Decision Tree Regression algorithms to forecast the skin permeability values of these compounds.

Results

When the Ridge regression technique was applied to the modified Wilschut et al dataset, the mean squared error was 0.20, the mean absolute error was 0.33 and the coefficient of determination (R²) was 0.61. The application of the same technique to the modified Stevens et al. dataset resulted in a mean squared error of 1.04, a mean absolute error of 0.5172 and an R-squared value of 0.18. The utilization of the Ridge regression method on the modified Steinmetz et al. dataset resulted in a mean squared error of 0.65, a mean absolute error of 0.67 and an R-squared Score of 0.48. When the Elasticnet regression approach was used on the modified Wilschut et al, the mean squared error was 0.24 and the coefficient of determination (R²) was 0.60. The utilization of the Elasticnet regression technique on the modified Steinmetz et al dataset led to a mean squared error of 0.88 and the coefficient of determination (R²) of 0.30. In comparison, Elasticnet regression technique on the modified Stevens et al dataset led to a mean squared error of 0.32 and the coefficient of determination (R²) of 0.42. The utilization of the Decision Tree(DT) regression on the modified Wilschut et al. dataset, resulted in the mean squared error of 0.28 and the coefficient of determination (R²) of 0.53. Decision Tree regression technique on the modified Stevens et al. dataset yielded a mean squared error: 0.31 and an R-squared Score :of 0.66. When the DT regression method was used on the Steinmetz et al dataset, the mean squared error was 0.89 and the R-squared Score was 0.14.

Conclusion

Our comparison analysis utilizing ElasticNet, Ridge, and Decision Tree regression models to forecast skin permeability across three datasets provides significant insights into the relationship between data quality and model efficacy. This finding is consistent with and enhances the advancing field of computer modeling in cutaneous absorption, specifically in medication development, cosmetic safety, and regulatory science.