Quantitative systems pharmacology (QSP) models have emerged as useful tools for evaluating the efficacy of Alzheimer's disease (AD) therapies. Bringing together a clinical focus with the mechanistic detail of systems biology, QSP models are well suited to the complexity of AD and have been used to predict treatment outcomes and support regulatory submissions. Therapies targeting the amyloid pathway are prominent in the AD clinical trial landscape, with anti-amyloid monoclonal antibodies representing the first approved disease-modifying therapies. To inform and facilitate future QSP model development, a systematic review of published QSP models focused on amyloid-targeting therapies for AD was completed. The PubMed and Web of Science databases were searched on February 1, 2025, identifying 540 candidate publications. Predefined exclusion and inclusion criteria were applied to identify seven published AD QSP models used to simulate treatment effects for one or more anti-amyloid therapies. The structure, development, and predictions of the models were summarized. Shared and contrasting model features were identified across included models. A set of model quality features was scored against a checklist of 15 criteria adapted from "best practice" guidelines for QSP. Model quality scores were generally low, ranging from 40% to 53%. Key quality issues related to model validation and reproducibility were identified; in particular, none of the seven papers provided executable model code. This systematic review provides useful context to support ongoing efforts to develop and refine QSP models such that they may better inform therapeutic strategies for the treatment of AD.
Chronic hepatitis B virus (HBV) infection remains a significant global health challenge. While the dynamic interplay between viral replication and host immune responses determines infection outcomes, the mechanisms driving the resolution of acute infection versus the emergence of chronicity remain incompletely understood. To address this challenge, we developed a detailed quantitative systems pharmacology (QSP) model of acute HBV infection capturing several key host immune and viral mechanisms absent in previous models. The model was parameterized using publicly available data and calibrated against clinical time-course datasets from multiple acute HBV case studies. Perturbation and local sensitivity analyses identified key drivers of biomarker dynamics, particularly hepatitis B virus DNA (HBV DNA), hepatitis B surface antigen (HBsAg), and alanine aminotransferase (ALT). These dynamics were most sensitive to parameters governing viral replication (e.g., HBV entry via the sodium taurocholate cotransporting polypeptide [NTCP] receptor, covalently closed circular DNA [cccDNA] formation, and hepatocyte turnover) and adaptive immune responses (e.g., CD8+ T cell activity, dendritic cell-mediated priming, and regulatory T cell [Treg]-driven immunosuppression). These influential parameters were used to generate a virtual population that reproduced the observed heterogeneity in biomarker trajectories. Notably, the magnitude and timing of biomarker peaks captured most of the variability, reflecting interindividual differences in individual immune responses and viral dynamics. While the current model nicely captures processes associated with acute HBV infections, it will be extended to different stages of chronic HBV with the objective of informing the rational design of novel therapies and supporting the development of curative HBV strategies.
The unprecedented effort to cope with the COVID-19 pandemic has unlocked the potential of mRNA vaccines as a powerful technology, set to become increasingly pervasive in the years to come. As in other areas of drug development, mathematical modeling is a pivotal tool to support and expedite the mRNA vaccine development process. This study introduces a Quantitative Systems Pharmacology (QSP) model that captures key immune responses following mRNA vaccine administration, encompassing both tissue-level and molecular-level events. The model mechanistically describes the biological processes from the uptake of mRNA by antigen-presenting cells at the injection site to the subsequent release of antibodies into the bloodstream. This two-layer model represents a first attempt to link the molecular mechanisms leading to antigen expression with the immune response, paving the way for the future integration of specific vaccine attributes, such as mRNA sequence features and nanotechnology-based delivery systems. Calibrated specifically for the BNT162b2 SARS-CoV-2 vaccine, the model has undergone successful validation across various dosing regimens and administration schedules. The results underscore the model's effectiveness in optimizing dosing strategies and highlighting critical differences in immune responses, particularly among low-responder groups such as the elderly. Furthermore, the model's adaptability has been demonstrated through its calibration for other mRNA vaccines, such as the Moderna mRNA-1273 vaccine, emphasizing its versatility and broad applicability in mRNA vaccine research and development.
Developed at Bayer Technology Services, PK-Sim and MoBi transitioned into the Open Systems Pharmacology (OSP) Suite, released as free open-source software in 2017. An active community with stakeholders from academia, industries, and regulators contributes to the continuous improvement of open-source model-informed drug development (MIDD). This perspective summarizes the latest advancements presented at the second OSP Community Conference (OSP-CC) hosted from 29 to 30th of September 2025 at Sanofi Paris, that gathered over 100 attendees from more than 40 institutions.
Quantitative Systems Pharmacology (QSP) is a powerful approach to provide decision-making support throughout the drug development process. QSP comes with many challenges in model development, validation, and applications. Traditional QSP workflows are limited by slow knowledge integration, labor-intensive model construction, inconsistent validation practices, and restricted scalability. In this work, we introduce QSP-Copilot, the first end-to-end AI-augmented solution designed to improve QSP modeling workflows by integrating a multi-agent system utilizing large language models (LLMs). QSP-Copilot provides modular support from project scoping and model structuring to model evaluation and reporting. Through the automation of routine tasks, QSP-Copilot reduces model development time by approximately 40% and improves methodological transparency through systematic documentation of literature sources and modeling assumptions. We demonstrate QSP-Copilot's application for two rare diseases of blood coagulation and Gaucher disease. In the blood coagulation case, automated extraction from ten peer-reviewed articles yielded 179 biological entity interaction pairs; out of these, only 105 unique mechanisms were retained after standardization. For Gaucher disease, screening nine articles produced 151 pairs, which were consolidated into 68 distinct biological interactions following the same post-processing workflow. The extraction precision for blood coagulation and Gaucher disease is 99.1% and 100.0%, respectively. QSP-Copilot extractions can be incorporated into effect diagrams with minimal expert filtering, significantly reducing the manual curation burden. The integration of AI-augmented workflows like QSP-Copilot represents a pivotal shift toward enhanced scalability and impact for QSP across the drug development pipelines, especially in disease areas where biological knowledge is sparse, such as rare diseases.
暂无摘要（点击查看详情）
Congenital thrombotic thrombocytopenic purpura (cTTP) is an ultra-rare, life-threatening thrombotic microangiopathy caused by a severe inherited deficiency of ADAMTS13, a von Willebrand factor (VWF) cleaving enzyme. Inadequate clinical endpoint data often make it challenging to statistically power clinical trials in ultra-rare diseases. Therefore, utilizing in vitro, adamts13-knockout mouse, literature-based, and clinical data, a quantitative systems pharmacology (QSP) model was developed to describe the mechanistic relationship between ADAMTS13, VWF, and platelet count, and to supplement evidence from clinical trials of recombinant ADAMTS13 (rADAMTS13) for the treatment of cTTP. The effect of long-term prophylaxis with rADAMTS13 versus plasma-based therapies (PBT) on platelet count in patients with cTTP was investigated. One-year clinical trial simulations of thrombocytopenia occurrences in 1000 virtual patients, phenotype-matched to a cTTP Phase 3 study population (NCT03393975), were produced. Simulations suggested that once-weekly (Q1W) or once every 2 weeks (Q2W) rADAMTS13 administered over 1 year resulted in fewer patients experiencing thrombocytopenia versus patients treated with PBT (e.g., Q2W [rADAMTS13] relative to Q2W [PBT], HR = 0.47 [platelet count drop to < 150 × 109/L], HR = 0.41 [< 100 × 109/L]). These results provide confirmative evidence to support the use of rADAMTS13 in cTTP by integrating the current mechanistic understanding of interactions between ADAMTS13 and VWF multimers as its substrate, as well as key downstream parameters, primarily platelet count. Virtual patient clinical simulations from the QSP model supported the regulatory approval of rADAMTS13 in cTTP, highlighting the significant potential of QSP modeling to supplement clinical trial data in rare disease drug development.
In 2017, the free and open-source software Open Systems Pharmacology (OSP) was launched. Since then, OSP has evolved from a small community into a diverse network of stakeholders committed to advancing open-source solutions for model-informed drug development (MIDD). In this context, the first OSP Community Conference was hosted by Novartis in Basel, Switzerland, on October 7-8, 2024, which gathered over 100 attendees from more than 40 institutions. This perspective synthesizes key insights from the conference.
The effectiveness of Bruton tyrosine kinase (BTK) inhibitors is influenced by the level of BTK occupancy in target tissues. In randomized phase 3 studies, progression-free survival (PFS) with zanubrutinib was superior to ibrutinib, whereas acalabrutinib was noninferior to ibrutinib in previously treated chronic lymphocytic leukemia. To establish a link between numerical differences in BTK occupancy and differentiated efficacy profiles among three covalent BTK inhibitors, quantitative systems pharmacology (QSP) modeling was employed. The model was developed to describe available clinical BTK occupancy data in patients with B-cell malignancies. Simulations of BTK occupancy were conducted for various clinical scenarios (e.g., dose interruption) and for bone marrow (BM), for which routine measurements are difficult. This model describes pharmacokinetics of BTK inhibitors; intracellular concentration of BTK inhibitors in peripheral blood mononuclear cells (PBMCs), BM, and lymph nodes (LNs); binding of BTK inhibitors with BTK; and BTK turnover rate. The model was validated using available clinical BTK occupancy data. Consistent with observed clinical data, the model predicted that zanubrutinib 160 mg twice daily resulted in higher median trough BTK occupancy in PBMCs, LNs, and BM compared with ibrutinib and acalabrutinib. Although the BTK occupancy differences at trough were relatively small between the BTK inhibitors, the differences were more pronounced after dose interruption. The current work underscores the importance of maintaining high BTK occupancy at steady-state trough and during treatment interruption to ensure maximal efficacy and provides an example of combining in vitro and clinical data to model receptor occupancy in tissues where measurements are challenging.
Neurofilament proteins are important constituents of neuronal cytoskeleton, along with microtubules. An increased concentration of neurofilament light (NfL) protein in cerebrospinal fluid (CSF) and plasma is considered a potential biomarker of axonal degeneration, which occurs in various neurodegenerative diseases including Alzheimer's disease (AD). The goal of this study was to develop a QSP model describing the change in the concentration of NfL in the brain, CSF, and plasma during the progression of AD for populations of AD patients manifesting different combinations of biomarkers (amyloid, tau, brain atrophy), to estimate the contributions of different mechanisms to neurodegeneration. The model correctly describes the dynamics of neurofilament proteins during neurodegeneration processes, which depend on cytoskeletal degradation and the release of neurofilament proteins from degenerated axons into cerebrospinal fluid and plasma. These processes are driven by disruptions of neuron homeostasis in AD, such as changes in protein degradation, axonal transport deficits, and the accumulation of pathological amyloid and hyperphosphorylated tau. The model was validated against clinical data and demonstrated correct predictions for anti-tau therapy while showing a tendency to overestimate efficacy of anti-amyloid therapy (lecanemab). This supports the idea that amyloid therapy contribution to neurodegeneration is limited, and that treatment should focus on other mechanisms.
The optimal post-transplant cyclophosphamide (PTCy) dose to prevent graft-versus-host disease (GVHD) after allogeneic hematopoietic cell transplant (HCT) is undefined. Data from a novel murine HLA-haploidentical HCT model suggested that PTCy has a dose-dependent effect associated with its efficacy in preventing GVHD, including reduced proliferation of conventional CD4+Foxp3- T-helper cells (Th) at Day 7, followed by the preferential expansion of regulatory CD4+CD25+Foxp3+ T cells (Tregs) at Day 21 after HCT. To better quantify the impact of the PTCy dose, we built a hybrid quantitative systems pharmacology (hybrid systems) model consisting of: (1) simulated pharmacokinetics of PTCy and its metabolites; (2) the donor immune system; and (3) a neural network mapping drug exposure and T-cell profiles to clinical scores of acute GVHD. Data were collected from murine MHC-haploidentical HCT studies of PTCy from 0 to 100 mg/kg/day on Days 3 and 4 after transplantation of donor cells. The final model successfully captured time-varying changes in donor Tcytotoxic CD8+ T cells (Tc), Th, and Tregs up to Day 21. The clinical score for GVHD was associated with numerous descriptors, with the immediately preceding GVHD score, body weight, PTCy dose, and donor Treg concentration in the blood and in the liver having the greatest contribution. This hybrid systems model quantitatively captures the beneficial impact of PTCy on Tc, Th, Tregs, and acute GVHD after HCT. The mediation through PTCy and metabolite concentrations and Tc, Th, and Treg numbers is underlined, and the model may be further used to refine and improve PTCy regimens.
Patient-specific covariates are commonly incorporated in pharmacometric and quantitative system pharmacology models to predict differences in pharmacokinetic or pharmacodynamic profiles between patients. When simulating new virtual populations of patients, generating realistic covariate sets that accurately reflect the correlation structures among covariates is essential to obtain reliable simulation outcomes. Copulas are joint distribution functions that characterize the dependence structures of patient covariates and enable the simulation of virtual populations. The current tutorial provides a step-by-step guide for understanding the concept of copulas and an overview of applications of copulas in pharmacometric research.
Trained pharmacometricians remain scarce in Africa due to limited training opportunities, lack of a pharmaceutical product development ecosystem, and emigration to high-income countries. The Applied Pharmacometrics Training (APT) fellowship program was established to address these gaps and specifically foster job creation for talent retention. We review the APT program's progress over 3 years and encourage collaboration to enhance local clinical data analysis in Africa. Initiated in 2021 by Pharmacometrics Africa, a non-profit educational entity, with support from partners including the Bill & Melinda Gates Foundation and Certara, the APT program targets African doctoral-level scientists and clinicians. This 6-month program is jointly managed by partners, with Pharmacometrics Africa handling logistics and sponsor liaison. Job creation initiatives include inviting fellows to join consulting teams or local research centers. Over the 3 year reporting period, 177 applications were received, with 27 individuals (41% female, median age 35 years) from nine African countries selected into and completing the full program. The fellows worked on 13 data analysis projects, with six so far being presented at international conferences and/or submitted for publication in peer-reviewed journals. Nine fellows have joined consulting teams or research centers working from offices in Africa. Currently, in the 3rd year, the APT program has demonstrated success in skills development, job creation, and fostering a critical mass of African pharmacometricians. Collaboration is essential for the sustainable advancement of model-informed drug development in Africa.
Delays in biological systems are a common phenomenon. The models for delays require specialized mathematical and numerical techniques such as transit compartments, delay differential equations (DDEs), and distributed DDEs (DDDEs). Because of mathematical complexity, DDEs and particularly DDDEs are infrequently used for modeling. DDEs are supported by most pharmacometric programs. Recently, DDDEs have been implemented in NONMEM that greatly improve the applicability of this technique in pharmacokinetic and pharmacodynamic (PKPD) modeling. The objective of this tutorial is to provide examples of PKPD models with delays and demonstrate how to implement them in NONMEM. All examples provide a brief description of the biology and pharmacology underlying model equations, explain how they are coded in the NONMEM control stream, and discuss results of data analysis models were used for. NONMEM codes for all models are presented in supporting information (Data S1). The tutorial concludes with a discussion of the pros and cons of presented delay modeling techniques with guidelines for which one might be preferred given the nature of the delay, available data, and the task to be performed.
Optimizing antibiotic therapy requires a holistic bench-to-bedside approach with interdisciplinary collaboration between pharmacologists, clinicians, microbiologists, and computational scientists. Novel experimental models provide insights into drug-pathogen interactions within complex host environments, while multiomics data provide details of the molecular mechanisms shaping bacterial responses. Pharmacometrics and machine learning can be used to integrate these insights into in silico models. This perspective highlights how these approaches-when used effectively and often together to build a systems-level view-can inform drug development and improve clinical decision-making, ensuring the right drug is given to each patient at the right time, at the right dose, and for the right duration.
For the treatment of Type 2 Diabetes, high efficacy approaches such as Glucagon-like peptide 1 (GLP-1)-based therapies are recommended for glucose control. Prediction of the clinical outcome of these therapies on glucose and hemoglobin A1c (HbA1c), using early available pharmacokinetic and in vitro efficacy information, can be a valuable tool for compound selection and supporting drug development. Our previously developed glucose homeostasis model (the 4GI model) is a systems model that is able to quantify drug effects on glucose based on in vitro potency and PK information. In this research, the model was coupled to an existing integrated glucose-red blood cell-HbA1c (IGRH) model for predicting the effects of GLP-1 and GLP-1/glucagon (dual) receptor agonists, liraglutide and cotadutide, on glucose and HbA1c. The 4GI model was validated for predicting 24-h glucose (Cglc,av) with minimal model calibration using short-term Ph2a continuous glucose monitoring (CGM) data. Subsequently, the predicted Cglc,av served as input for the HbA1c model to assess the predictiveness of the combined 4GI-HbA1c model on HbA1c. The resulting combined model was used in cotadutide's clinical development by providing predictive insights into the 26 weeks glucose and HbA1c dynamics of the Ph2b study prior to its initiation. Retrospective analysis showed that the model adequately predicted the effect of cotadutide and liraglutide on fasting plasma glucose and HbA1c (Root Means Square Percent Error (RMSPE) 5.9% and 13%, respectively). This demonstrates the potential of the 4GI-HbA1c systems model as a valuable tool in supporting the clinical development of novel GLP-1 and/or glucagon agonists.
Acetaminophen (APAP) has been formulated as immediate-, modified-, and extended-release tablets (APAP-IR, -MR, and -ER, respectively). However, there was concern that APAP-MR previously available in Europe could form a bezoar after a large overdose, leading to delayed absorption and atypical pharmacokinetics (PK) compared to APAP-IR, and that current treatment guidelines developed for APAP overdose to prevent severe hepatotoxicity are inappropriate for APAP-MR. In contrast, APAP-ER caplets available in the United States are designed with an IR layer and an erodible ER layer. Using modeling and simulation, predicted PK and hepatotoxicity biomarkers following various acute overdose and repeated supratherapeutic ingestion (RSTI) scenarios with APAP-IR and APAP-ER were compared to investigate the differences between these two formulations. The existing APAP-IR representation within DILIsym v8A, a quantitative systems toxicology model of drug-induced liver injury, was updated, and an APAP-ER model was developed, using newly acquired in vitro (e.g., tiny-TIMsg) and clinical data. The model and simulated populations (SimPops) representing healthy adults were extensively validated, before simulating PK and three clinically useful hepatic biomarkers after various overdose scenarios. On average, APAP exposure after acute overdose and RSTI in healthy adults was predicted to be slightly lower for APAP-ER compared to APAP-IR, partially due to lower APAP absorption for APAP-ER, while not markedly impacting the expected time course of APAP plasma concentrations. Similar hepatic biomarker profiles were predicted for both APAP formulations. Based on these results, the APAP overdose consensus treatment guidelines updated in 2023 are not further impacted by this report.
Acetaminophen (APAP), an over-the-counter analgesic and antipyretic, can cause hepatotoxicity when ingested in large overdoses. APAP has multiple formulations including immediate-release (IR) and extended-release (ER) preparations. A recently published consensus statement on the management of APAP poisoning indicated that management of APAP-ER overdose is the same as that for APAP-IR overdose. Consistent with this consensus, it was previously reported that quantitative systems toxicology (QST) modeling using DILIsym predicted similar pharmacokinetic (PK) and hepatic biomarker profiles for the APAP-ER and APAP-IR formulations after overdose in healthy adults. Hepatic injury from APAP is caused by the reactive metabolite, N-acetyl-ρ-benzoquinone imine (NAPQI), which is formed predominantly by CYP2E1-mediated metabolism and eliminated by hepatic glutathione. As such, conditions that can increase NAPQI production (e.g., CYP2E1 induction by alcohol) or decrease hepatic glutathione stores (e.g., underling liver disease) may impact PK and susceptibility to hepatotoxicity after overdose of APAP-IR and APAP-ER. In the current study, APAP-IR and APAP-ER models in chronic alcohol users and individuals with low hepatic glutathione were developed and verified within DILIsym. Simulations using verified models predicted similar PK and hepatic biomarker profiles for the APAP-ER and APAP-IR formulations in moderate and excessive chronic alcohol users and adults with low hepatic glutathione levels after single acute overdoses up to ~100 g and repeat supratherapeutic ingestions (up to 7.8 g/day for 10 days). These results further support that approaches to manage APAP-IR overdoses can be applied to manage APAP-ER overdoses in adults with chronic alcohol consumption or lower hepatic glutathione levels.
Doxorubicin (DOX) and trastuzumab (TmAb) are widely used to treat HER2-positive breast cancer (BC), as monotherapies and in combination (DOX + TmAb). While highly effective, their combined use significantly increases the risk of irreversible cardiotoxicity, posing a major clinical concern. B-type natriuretic peptide (BNP) and NT-proBNP are serum biomarkers of early cardiotoxicity. Understanding the dynamic relationship between these biomarkers and intracellular apoptosis pathways is key to predicting and mitigating treatment-induced cardiotoxicity. This study aims to extend a previously developed multiscale modeling framework of DOX-induced cardiotoxicity to include DOX + TmAb combinatorial effects and to predict clinical outcomes. Human cardiomyocytes were exposed to different concentrations of DOX, TmAb, DOX + TmAb, or control for 96 h. Time-course data for caspase-9 and -3 expression, cell viability, and BNP were collected and used to develop mathematical models for intracellular apoptosis-signaling protein dynamics, cardiomyocyte viability, and cardiomyocyte injury biomarkers. The cellular model was scaled up to humans with a previously published TmAb human PBPK model using NT-proBNP data and evaluated with left ventricular ejection fraction measurements. The quantitative systems toxicology (QST) model successfully captured in vitro dynamic data across treatment groups. Caspase-3 drove the cardiomyocyte-death model. Multiplicative and additive relationships characterized drug interactions to reflect the enhanced cardiotoxicity seen with DOX + TmAb. The predicted clinical BNP changes were consistent with LVEF dynamics from BC patients treated with TmAb. The QST-PBPK model bridges in vitro experimental findings with clinical cardiotoxicity outcomes. It provides a predictive tool for cardiotoxicity, aiding potentially in dose optimization and clinical monitoring for HER2-positive BC patients.
This year marks the 20th edition of the Basel Modeling and Simulation (M&S) Seminar, an initiative rooted in a commitment to promoting the exchange of the latest advancements in pharmacometrics and related disciplines in the region of Basel, Switzerland. This article provides insight into the history of this event, its operations to the present date, and a glimpse at the future.