Flexible Web service infrastructure for the development and deployment of predictive models

Area of Science:

• Computational chemistry
• Bioinformatics
• Statistical modeling

Background:

• Predictive statistical models are crucial in drug design.
• Traditional deployment via web pages limits model accessibility and integration.
• A need exists for more flexible and generalizable model deployment strategies.

Purpose of the Study:

• To present a novel web service infrastructure for deploying predictive statistical models.
• To offer a flexible and generalizable approach to model deployment in drug design.
• To demonstrate the utility of the proposed infrastructure using random forest models.

Main Methods:

• Development of a web service infrastructure using the R programming language.
• Implementation of a flexible deployment strategy for predictive models.
• Application of random forest models to two distinct datasets for demonstration.

Main Results:

• The R-based web service infrastructure allows diverse access methods, including web pages and workflow tools.
• The proposed approach overcomes limitations of traditional web page deployments.
• Successful deployment of random forest models was achieved for two datasets.

Conclusions:

• The developed web service infrastructure provides a flexible and generalizable solution for deploying predictive models in drug design.
• This approach enhances model accessibility and facilitates integration with various tools.
• The R-based infrastructure offers a robust platform for advancing computational drug discovery.