CBK model composition using paired web services and executable functions: A demonstration for individualizing...

Area of Science:

• Biomedical Informatics
• Computational Biology
• Health Systems Science

Background:

• Learning health systems require integrated computable biomedical knowledge (CBK) models.
• Existing methods for combining CBK models are often complex and non-standardized.
• There is a need for more efficient and modular approaches to CBK model composition.

Purpose of the Study:

• To present a novel method for composing CBK models using Knowledge Objects and World Wide Web (WWW) technologies.
• To demonstrate the feasibility of creating standardized and reusable composite CBK models.
• To facilitate the integration of diverse CBK models within learning health systems.

Main Methods:

• Utilized compound digital objects called Knowledge Objects, encapsulating CBK models with metadata and runtime requirements.
• Developed the KGrid Activator tool to instantiate CBK models within open-source runtimes and expose them via RESTful APIs.
• Established a CBK model composition method by interconnecting model inputs and outputs through the KGrid Activator.

Main Results:

• Successfully developed a complex composite CBK model (CM-IPP) from 42 submodels to estimate life-gain.
• Created an externalized, highly modularized CM-IPP implementation runnable in standard server environments.
• Demonstrated the feasibility of composing CBK models using Knowledge Objects and distributed computing.

Conclusions:

• CBK model composition is feasible using Knowledge Objects and distributed computing technologies.
• The proposed method supports the creation of large ecosystems of adaptable and reusable CBK models.
• Future work should focus on optimizing model boundaries and submodel organization for enhanced reusability.