Partial identification of virtual population predictions in quantitative systems pharmacology

Calibration of a quantitative systems pharmacology (QSP) model to population summaries need not identify its predictions under a new intervention. We formulate prediction extrema over calibration-compatible virtual populations as a generalized moment problem and implement finite-cohort linear programming with support-guided cohort expansion. The approach yields explicit alternative populations and constructive inner approximations to the identified prediction interval; it does not compute globally certified outer bounds. We apply it to a published atopic dermatitis model with 51 varying parameters and 14 state variables. Calibration targets are generated from a truncated reference distribution based on the published parameter distributions, rather than extracted from clinical trials. In the primary run, 60,400 candidate patients and 90 moment constraints admit mean week-16 EASI improvement of 62.1–94.5% under hypothetical IL-13 plus IL-22 blockade, compared with 79.5% in the reference population. The corresponding responder-rate range is 31.4–100.0%, compared with 72.5%. Under a one-sided weight cap of twice the empirical reference weights, the responder range is 58.1–86.8%. A second complete run supports the same qualitative conclusion. Wider prediction ranges constitute constructive evidence of non-identification, whereas narrow ranges found by a heuristic search cannot establish identification. This computational case study demonstrates how explicit calibration-compatible populations can reveal prediction uncertainty beyond a selected reference population. Clinical calibration, comprehensive plausibility assessment and global optimization certificates remain future work.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-12
DOI
https://doi.org/10.5281/zenodo.24176886
Primary Topic
Statistical Methods in Clinical Trials
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Partial identification of virtual population predictions in quantitative systems pharmacology

Eva Paunova
Zenodo (CERN European Organization for Nuclear Research)
Statistical Methods in Clinical Trials
preprint

Partial identification of virtual population predictions in quantitative systems pharmacology

Eva Paunova
preprint en

Abstract

Calibration of a quantitative systems pharmacology (QSP) model to population summaries need not identify its predictions under a new intervention. We formulate prediction extrema over calibration-compatible virtual populations as a generalized moment problem and implement finite-cohort linear programming with support-guided cohort expansion. The approach yields explicit alternative populations and constructive inner approximations to the identified prediction interval; it does not compute globally certified outer bounds. We apply it to a published atopic dermatitis model with 51 varying parameters and 14 state variables. Calibration targets are generated from a truncated reference distribution based on the published parameter distributions, rather than extracted from clinical trials. In the primary run, 60,400 candidate patients and 90 moment constraints admit mean week-16 EASI improvement of 62.1–94.5% under hypothetical IL-13 plus IL-22 blockade, compared with 79.5% in the reference population. The corresponding responder-rate range is 31.4–100.0%, compared with 72.5%. Under a one-sided weight cap of twice the empirical reference weights, the responder range is 58.1–86.8%. A second complete run supports the same qualitative conclusion. Wider prediction ranges constitute constructive evidence of non-identification, whereas narrow ranges found by a heuristic search cannot establish identification. This computational case study demonstrates how explicit calibration-compatible populations can reveal prediction uncertainty beyond a selected reference population. Clinical calibration, comprehensive plausibility assessment and global optimization certificates remain future work.

Zenodo (CERN European Organization for Nuclear Research)
Good health and well-being
Statistical Methods in Clinical Trials
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon