Integrative Systems Pharmacology Modeling

Abstract

Drug development suffers a 90% clinical trial failure rate, costing an average $2.6 billion per approved drug, largely because preclinical models inadequately predict human pharmacology. Systems pharmacology -- integrating molecular target interactions, signalling pathway dynamics, cellular responses, organ-level physiology, and patient variability into quantitative computational models -- promises to bridge this translational gap. However, current approaches model individual scales in isolation: pharmacokinetic (PK) models predict drug concentration, pharmacodynamic (PD) models predict target engagement, and systems biology models predict pathway responses, without seamless cross-scale integration. This study developed and evaluated five pharmacology modelling approaches -- classical compartmental PK/PD, physiologically-based PK (PBPK), quantitative systems pharmacology (QSP), AI-driven PK/PD (neural ODE), and a proposed Multi-Scale Integrative Pharmacology Platform (MSIPP) combining PBPK-informed tissue concentrations, mechanistic QSP pathway models, and graph neural network-predicted drug-target-pathway interactions with virtual patient population simulation -- for predicting clinical efficacy and toxicity of three oncology drugs: osimertinib (EGFR inhibitor), pembrolizumab (anti-PD-1), and venetoclax (BCL-2 inhibitor). MSIPP achieved the highest prediction accuracy for clinical endpoints: tumour response rate predicted within 4.2 +- 2.8% of observed Phase III results (vs 18.6% for classical PK/PD, 12.4% for PBPK, 8.8% for QSP, 10.2% for neural ODE; p < 0.001), dose-limiting toxicity incidence within 3.8 +- 2.4% of observed, and optimal dose selection matching the approved dose in 3/3 drugs (vs 1/3 for classical PK/PD). Virtual clinical trials simulated by MSIPP reproduced the efficacy-toxicity profiles of actual Phase III trials with Pearson r = 0.94 across endpoints. These results establish multi-scale integrative systems pharmacology as a transformative approach for drug development decision-making