From Data to Parameters: FIVEDB as a Tool for Efficient QSP Model Calibration

Objectives: Parameter calibration in Quantitative Systems Pharmacology (QSP) models is an essential but time-consuming task, involving literature analysis, extraction of relevant in vitro data, and their application to parameter estimation. To facilitate this workflow, we developed FIVEDB [ 1], a manually curated database that aggregates experimental data specifically from human samples, standardizes data units, and computes kinetic parameters describing cell death, differentiation, migration, proliferation, and secretion. The primary objectives of FIVEDB are: (1) to collect in vitro experimental data and translate them into kinetic parameters; (2) to integrate diverse literature data into a structured, queryable format; (3) to reduce the time spent on manual data extraction and parameter estimation. Methods: FIVEDB s an interactive web application built using ASP.NET MVC, Microsoft SQL Server, and Bootstrap. Key innovations include: Automatic Parameter Calculation: Data from time-course and dose-response experiments are processed to estimate parameters (e.g., kbase, Emax/Imax, EC50/IC50) using an InSysBio-developed approach [ 2]. Unit Standardization: Different types of input data are converted into standard units: pM, kcell/L, hours. Quality Control: Includes two-step manual validation and automated range-checking algorithms. Query-Based Retrieval: Filters (e.g., cell type, regulator, patient state) allow targeted searches. Results can be exported in a convenient CSV tabular format for further analysis. Dynamic Updates: Over 1,000 new values are added monthly Results: FIVEDB contains ~35,000 parameter values related to cell processes, curated from 1,835 publications. Two case studies illustrate its functionality: Basal cytokine production constants for various immune cell types were retrieved, visualized, and analyzed. The resulting cytokine profiles revealed a high level of IL-8 secretion in stimulated monocytes and neutrophils, along with pronounced secretion of IL-6 and CCL2 by monocytes. Modulation parameters were used to assess cytokine production in response to IFN-γ (a common secretion regulator in human cells). While not directly affecting processes, modulators influence activator/inhibitor efficacy. CXCL9 showed the strongest modulation effect, enhancing IFN-γ activity over 60-fold. Conclusions: FIVEDB is a user-friendly tool that integrates experimental in vitro data with derived kinetic parameters It supports QSP modelers by simplifying a labor-intensive stage of model calibration and minimizing time spent on manual data extraction. Through automated parameter calculation, standardized data formatting, and regular updates, FIVEDB provides a practical and effective solution for robust models calibration and development.