CrossLabFit: A novel framework for integrating qualitative and quantitative data across multiple labs for model...

Area of Science:

• Biomedical research
• Computational biology
• Data science

Background:

• Integrating computational models with experimental data is crucial for biomedical insights.
• Parameter fitting often requires extensive data, which is difficult to obtain from a single source.
• Existing methods struggle to harmonize diverse data types from multiple laboratories.

Purpose of the Study:

• To present a novel methodology, CrossLabFit, for integrating data from multiple laboratories.
• To overcome the limitations of single-lab data collection in computational modeling.
• To harmonize disparate qualitative assessments into a unified framework for parameter estimation.

Main Methods:

• Developed CrossLabFit to integrate data from multiple labs, including qualitative assessments.
• Utilized machine learning clustering to represent qualitative constraints as dynamic "feasible windows".
• Implemented a GPU-accelerated differential evolution algorithm for parameter estimation using integrated data.

Main Results:

• Demonstrated significant improvements in model accuracy and parameter identifiability through case studies.
• Successfully harmonized qualitative and quantitative data for robust parameter estimation.
• Validated the effectiveness of the "feasible windows" approach in guiding model adherence to experimental trends.

Conclusions:

• CrossLabFit offers a new paradigm for collaborative science by enabling data integration across laboratories.
• The methodology provides a roadmap for combining and comparing findings from different studies.
• This approach enhances the understanding of biological systems through more accurate and identifiable computational models.