Virtual Populations for Quantitative Systems Pharmacology Models

Synthetic Data in Healthcare and Drug Development: Definitions, Regulatory Frameworks, Issues

With the recent and evolving regulatory frameworks regarding the usage of Artificial Intelligence (AI) in both drug and medical device development, the differentiation between data derived from observed ('true' or 'real') sources and artificial data obtained using process-driven and/or (data-driven) algorithmic processes is emerging as a critical consideration in clinical research and regulatory discourse. We conducted a critical literature review that revealed evidence of the current ambivalent usage of the term "synthetic" (along with derivative terms) to refer to "true/observed" data in the context of clinical trials and AI-generated data (or "artificial" data). This paper, stemming from a critical evaluation of different perspectives captured from the scientific literature and recent regulatory endeavors, seeks to elucidate this distinction, exploring their respective utilities, regulatory stances, and upcoming needs, as well as the potential for both data types in advancing medical science and therapeutic development.

Circumventing glioblastoma resistance to temozolomide through optimal drug combinations designed by systems pharmacology and machine learning

BACKGROUND AND PURPOSE

Glioblastoma (GBM), the most frequent and aggressive brain tumour in adults, is associated with a dismal prognostic despite intensive treatment involving surgery, radiotherapy and temozolomide (TMZ)-based chemotherapy. The initial or acquired resistance of GBM to TMZ appeals for precision medicine approaches to the design of novel efficient combination pharmacotherapies. Such investigation needs to account for the overexpression of the O6-methylguanine-DNA methyl-transferase (MGMT) repair enzyme which is responsible for TMZ resistance in patients.

EXPERIMENTAL APPROACH

A comprehensive approach combining quantitative systems pharmacology (QSP) models and machine learning (ML) was undertaken to design TMZ-based drug combinations circumventing the initial resistance to the alkylating agent.

KEY RESULTS

A QSP model representing TMZ cellular pharmacokinetics-pharmacodynamics and dysregulated pathways in GBM was developed and validated using multi-type time- and dose-resolved datasets, available in control or MGMT-overexpressing cells. In silico drug screening and subsequent experimental validation identified a strategy to re-sensitise TMZ-resistant cells consisting in combining TMZ with inhibitors of the base excision repair and of homologous recombination. Using ML, functional signatures of response to such optimal multi-agent therapy were derived to assist decision-making in patients.

CONCLUSION AND IMPLICATIONS

We successfully demonstrated the relevance of combined QSP and ML to design efficient drug combinations re-sensitising glioblastoma cells initially resistant to TMZ. The developed framework may further serve to identify personalised therapies and administration schedules by extending it to account for additional patient-specific altered pathways and whole-body features.

A Framework for Quantitative Systems Pharmacology Model Execution

A mathematical model can be defined as a theoretical approximation of an observed pattern. The specific form of the model and the associated mathematical methods are typically dictated by the question(s) to be addressed by the model and the underlying data. In the context of research and development of new medicines, these questions often focus on the dose-exposure-response relationship.The general workflow for model development and application can be delineated in three major elements: defining the model, qualifying the model, and performing simulations. These elements may vary significantly depending on modeling objectives. Quantitative systems pharmacology (QSP) models address the formidable challenge of quantitatively and mechanistically characterizing human and animal biology, pathophysiology, and therapeutic intervention.QSP model development, by necessity, relies heavily on preexisting knowledge, requires a comprehensive understanding of current physiological concepts, and often makes use of heterogeneous and aggregated datasets from multiple sources. This reliance on diverse datasets presents an upfront challenge: the determination of an optimal model structure while balancing model complexity and uncertainty. Additionally, QSP model calibration is arduous due to data scarcity (particularly at the human subject level), which necessitates the use of a variety of parameter estimation approaches and sensitivity analyses, earlier in the modeling workflow as compared to, for example, population modeling. Finally, the interpretation of model-based predictions must be thoughtfully aligned with the data and the mathematical methods applied during model development.The purpose of this chapter is to provide readers with a high-level yet comprehensive overview of a QSP modeling workflow, with an emphasis on the various challenges encountered in this process. The workflow is centered around the construction of ordinary differential equation models and may be extended beyond this framework. It includes the fundamentals of systematic literature reviews, the selection of appropriate structural model equations, the analysis of system behavior, model qualification, and the application of various types of model-based simulations. The chapter concludes with details on existing software options suitable for implementing the described methodologies.This workflow may serve as a valuable resource to both newcomers and experienced QSP modelers, offering an introduction to the field as well as operating procedures and references for routine analyses.

A Generative Shape Compositional Framework to Synthesize Populations of Virtual Chimeras

Generating virtual organ populations that capture sufficient variability while remaining plausible is essential to conduct in silico trials (ISTs) of medical devices. However, not all anatomical shapes of interest are always available for each individual in a population. The imaging examinations and modalities used can vary between subjects depending on their individualized clinical pathways. Different imaging modalities may have various fields of view and are sensitive to signals from other tissues/organs, or both. Hence, missing/partially overlapping anatomical information is often available across individuals. We introduce a generative shape model for multipart anatomical structures, learnable from sets of unpaired datasets, i.e., where each substructure in the shape assembly comes from datasets with missing or partially overlapping substructures from disjoint subjects of the same population. The proposed generative model can synthesize complete multipart shape assemblies coined virtual chimeras (VCs). We applied this framework to build VCs from databases of whole-heart shape assemblies that each contribute samples for heart substructures. Specifically, we propose a graph neural network-based generative shape compositional framework, which comprises two components, a part-aware generative shape model that captures the variability in shape observed for each structure of interest in the training population and a spatial composition network that assembles/composes the structures synthesized by the former into multipart shape assemblies (i.e., VCs). We also propose a novel self-supervised learning scheme that enables the spatial composition network to be trained with partially overlapping data and weak labels. We trained and validated our approach using shapes of cardiac structures derived from cardiac magnetic resonance (MR) images in the UK Biobank (UKBB). When trained with complete and partially overlapping data, our approach significantly outperforms a principal component analysis (PCA)-based shape model (trained with complete data) in terms of generalizability and specificity. This demonstrates the superiority of the proposed method, as the synthesized cardiac virtual populations are more plausible and capture a greater degree of shape variability than those generated by the PCA-based shape model.

Virtual clinical trials via a QSP immuno-oncology model to simulate the response to a conditionally activated PD-L1 targeting antibody in NSCLC

Recently, immunotherapies for antitumoral response have adopted conditionally activated molecules with the objective of reducing systemic toxicity. Amongst these are conditionally activated antibodies, such as PROBODY® activatable therapeutics (Pb-Tx), engineered to be proteolytically activated by proteases found locally in the tumor microenvironment (TME). These PROBODY® therapeutics molecules have shown potential as PD-L1 checkpoint inhibitors in several cancer types, including both effectiveness and locality of action of the molecule as shown by several clinical trials and imaging studies. Here, we perform an exploratory study using our recently published quantitative systems pharmacology model, previously validated for triple-negative breast cancer (TNBC), to computationally predict the effectiveness and targeting specificity of a PROBODY® therapeutics drug compared to the non-modified antibody. We begin with the analysis of anti-PD-L1 immunotherapy in non-small cell lung cancer (NSCLC). As a first contribution, we have improved previous virtual patient selection methods using the omics data provided by the iAtlas database portal compared to methods previously published in literature. Furthermore, our results suggest that masking an antibody maintains its efficacy while improving the localization of active therapeutic in the TME. Additionally, we generalize the model by evaluating the dependence of the response to the tumor mutational burden, independently of cancer type, as well as to other key biomarkers, such as CD8/Treg Tcell and M1/M2 macrophage ratio. While our results are obtained from simulations on NSCLC, our findings are generalizable to other cancer types and suggest that an effective and highly selective conditionally activated PROBODY® therapeutics molecule is a feasible option.

In-silico trial emulation to predict the cardiovascular protection of new lipid-lowering drugs: an illustration through the design of the SIRIUS programme

INTRODUCTION

Inclisiran, an siRNA targeting hepatic PCSK9 mRNA, administered twice-yearly (after initial and 3-month doses), substantially and sustainably reduced LDL-cholesterol (LDL-C) in Phase III trials. Whether lowering LDL-C with inclisiran translates into a reduced risk of major adverse cardiovascular events (MACE) is not yet established. In-silico trials applying a disease computational model to virtual patients receiving new treatments allow to emulate large scale long-term clinical trials. The SIRIUS in-silico trial programme aims to predict the efficacy of inclisiran on CV events in individuals with established atherosclerotic cardiovascular disease (ASCVD).

METHODS AND RESULTS

A knowledge-based mechanistic model of ASCVD was built, calibrated, and validated to conduct the SIRIUS programme (NCT05974345) aiming to predict the effect of inclisiran on CV outcomes. The SIRIUS Virtual Population included patients with established ASCVD (previous myocardial infarction (MI), previous ischemic stroke (IS), previous symptomatic lower limb peripheral arterial disease (PAD) defined as either intermittent claudication with ankle-brachial index <0.85, prior peripheral arterial revascularization procedure, or vascular amputation) and fasting LDL-C ≥ 70 mg/dL, despite stable (≥4 weeks) well-tolerated lipid-lowering therapies.SIRIUS is an in-silico multi-arm trial programme. It follows an idealized crossover design where each virtual patient is its own control, comparing inclisiran to (i) placebo as adjunct to high-intensity statin therapy with or without ezetimibe, (ii) ezetimibe as adjunct to high-intensity statin therapy, (iii) evolocumab as adjunct to high-intensity statin therapy and ezetimibe.The co-primary efficacy outcomes are based on the time to the first occurrence of any component of 3P-MACE (composite of CV death, nonfatal MI, or nonfatal IS) and time to occurrence of CV death over 5 years.

PERSPECTIVES/CONCLUSION

The SIRIUS in-silico trial programme will provide early insights regarding potential effect of inclisiran on MACE in ASCVD patients, several years before the availability of the results from ongoing CV outcomes trials (ORION-4 and VICTORION-2-P).

CLINICAL TRIAL REGISTRATION

Clinicaltrials.gov identifier: NCT05974345.

An industry perspective on current QSP trends in drug development

2023 marks the 10th anniversary of Natpara's submission to the US FDA, which led to the first recorded regulatory interaction where a decision was supported by Quantitative and Systems Pharmacology (QSP) simulations. It had taken about 5 years for the timid QSP discipline to emerge as an effective Model-Informed Drug Development (MIDD) tool with visible impact in the pharmaceutical industry. Since then, the presence of QSP in the regulatory environment has continued to increase, to the point that the Agency reported 60 QSP submissions in 2020 alone, representing ~ 4% of their annual IND submissions [1]. What sort of industry mindset has enabled QSP to reach this level of success? How does QSP fit within the MIDD paradigm? Does QSP mean the same to Discovery and to Clinical Development projects? How do 'platforms' compare to 'fit-for-purpose' QSP models in an industrial setting? Can QSP and empirical Pharmacokinetic-Pharmacodynamic (PKPD) modelling be complementary? What level of validation is required to inform drug development decisions? This article reflects on all these questions, in particular addressing those audiences with limited line-of-sight into the drug industry decision-making machinery.

Integration of Clinical Trial Spatial Multiomics Analysis and Virtual Clinical Trials Enables Immunotherapy Response Prediction and Biomarker Discovery

Due to the lack of treatment options, there remains a need to advance new therapeutics in hepatocellular carcinoma (HCC). The traditional approach moves from initial molecular discovery through animal models to human trials to advance novel systemic therapies that improve treatment outcomes for patients with cancer. Computational methods that simulate tumors mathematically to describe cellular and molecular interactions are emerging as promising tools to simulate the impact of therapy entirely in silico, potentially greatly accelerating delivery of new therapeutics to patients. To facilitate the design of dosing regimens and identification of potential biomarkers for immunotherapy, we developed a new computational model to track tumor progression at the organ scale while capturing the spatial heterogeneity of the tumor in HCC. This computational model of spatial quantitative systems pharmacology was designed to simulate the effects of combination immunotherapy. The model was initiated using literature-derived parameter values and fitted to the specifics of HCC. Model validation was done through comparison with spatial multiomics data from a neoadjuvant HCC clinical trial combining anti-PD1 immunotherapy and a multitargeted tyrosine kinase inhibitor cabozantinib. Validation using spatial proteomics data from imaging mass cytometry demonstrated that closer proximity between CD8 T cells and macrophages correlated with nonresponse. We also compared the model output with Visium spatial transcriptomics profiling of samples from posttreatment tumor resections in the clinical trial and from another independent study of anti-PD1 monotherapy. Spatial transcriptomics data confirmed simulation results, suggesting the importance of spatial patterns of tumor vasculature and TGFβ in tumor and immune cell interactions. Our findings demonstrate that incorporating mathematical modeling and computer simulations with high-throughput spatial multiomics data provides a novel approach for patient outcome prediction and biomarker discovery. Significance: Incorporating mathematical modeling and computer simulations with high-throughput spatial multiomics data provides an effective approach for patient outcome prediction and biomarker discovery.

Assessing the Role of Patient Generation Techniques in Virtual Clinical Trial Outcomes

Virtual clinical trials (VCTs) are growing in popularity as a tool for quantitatively predicting heterogeneous treatment responses across a population. In the context of a VCT, a plausible patient is an instance of a mathematical model with parameter (or attribute) values chosen to reflect features of the disease and response to treatment for that particular patient. A number of techniques have been introduced to determine the set of model parametrizations to include in a virtual patient cohort. These methodologies generally start with a prior distribution for each model parameter and utilize some criteria to determine whether a parameter set sampled from the priors should be included or excluded from the plausible population. No standard technique exists, however, for generating these prior distributions and choosing the inclusion/exclusion criteria. In this work, we rigorously quantify the impact that VCT design choices have on VCT predictions. Rather than use real data and a complex mathematical model, a spatial model of radiotherapy is used to generate simulated patient data and the mathematical model used to describe the patient data is a two-parameter ordinary differential equations model. This controlled setup allows us to isolate the impact of both the prior distribution and the inclusion/exclusion criteria on both the heterogeneity of plausible populations and on predicted treatment response. We find that the prior distribution, rather than the inclusion/exclusion criteria, has a larger impact on the heterogeneity of the plausible population. Yet, the percent of treatment responders in the plausible population was more sensitive to the inclusion/exclusion criteria utilized. This foundational understanding of the role of virtual clinical trial design should help inform the development of future VCTs that use more complex models and real data.

Using virtual patient cohorts to uncover immune response differences in cancer and immunosuppressed COVID-19 patients

The COVID-19 pandemic caused by the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) resulted in millions of deaths globally. Adults with immunosuppression (e.g., solid organ transplant recipients) and those undergoing active cancer treatments experience worse infections and more severe COVID-19. It is difficult to conduct clinical studies in these populations, resulting in a restricted amount of data that can be used to relate mechanisms of immune dysfunction to COVID-19 outcomes in these vulnerable groups. To study immune dynamics after infection with SARS-CoV-2 and to investigate drivers of COVID-19 severity in individuals with cancer and immunosuppression, we adapted our mathematical model of the immune response during COVID-19 and generated virtual patient cohorts of cancer and immunosuppressed patients. The cohorts of plausible patients recapitulated available longitudinal clinical data collected from patients in Montréal, Canada area hospitals. Our model predicted that both cancer and immunosuppressed virtual patients with severe COVID-19 had decreased CD8+ T cells, elevated interleukin-6 concentrations, and delayed type I interferon peaks compared to those with mild COVID-19 outcomes. Additionally, our results suggest that cancer patients experience higher viral loads (however, with no direct relation with severity), likely because of decreased initial neutrophil counts (i.e., neutropenia), a frequent toxic side effect of anti-cancer therapy. Furthermore, severe cancer and immunosuppressed virtual patients suffered a high degree of tissue damage associated with elevated neutrophils. Lastly, parameter values associated with monocyte recruitment by infected cells were found to be elevated in severe cancer and immunosuppressed patients with respect to the COVID-19 reference group. Together, our study highlights that dysfunction in type I interferon and CD8+ T cells are key drivers of immune dysregulation in COVID-19, particularly in cancer patients and immunosuppressed individuals.

From virtual patients to digital twins in immuno-oncology: lessons learned from mechanistic quantitative systems pharmacology modeling

Virtual patients and digital patients/twins are two similar concepts gaining increasing attention in health care with goals to accelerate drug development and improve patients' survival, but with their own limitations. Although methods have been proposed to generate virtual patient populations using mechanistic models, there are limited number of applications in immuno-oncology research. Furthermore, due to the stricter requirements of digital twins, they are often generated in a study-specific manner with models customized to particular clinical settings (e.g., treatment, cancer, and data types). Here, we discuss the challenges for virtual patient generation in immuno-oncology with our most recent experiences, initiatives to develop digital twins, and how research on these two concepts can inform each other.

Applying quantitative and systems pharmacology to drug development and beyond: An introduction to clinical pharmacologists

ABSTRACT

Quantitative and systems pharmacology (QSP) is an innovative and integrative approach combining physiology and pharmacology to accelerate medical research. This review focuses on QSP's pivotal role in drug development and its broader applications, introducing clinical pharmacologists/researchers to QSP's quantitative approach and the potential to enhance their practice and decision-making. The history of QSP adoption reveals its impact in diverse areas, including glucose regulation, oncology, autoimmune disease, and HIV treatment. By considering receptor-ligand interactions of various cell types, metabolic pathways, signaling networks, and disease biomarkers simultaneously, QSP provides a holistic understanding of interactions between the human body, diseases, and drugs. Integrating knowledge across multiple time and space scales enhances versatility, enabling insights into personalized responses and general trends. QSP consolidates vast data into robust mathematical models, predicting clinical trial outcomes and optimizing dosing based on preclinical data. QSP operates under a "learn and confirm paradigm," integrating experimental findings to generate testable hypotheses and refine them through precise experimental designs. An interdisciplinary collaboration involving expertise in pharmacology, biochemistry, genetics, mathematics, and medicine is vital. QSP's utility in drug development is demonstrated through integration in various stages, predicting drug responses, optimizing dosing, and evaluating combination therapies. Challenges exist in model complexity, communication, and peer review. Standardized workflows and evaluation methods ensure reliability and transparency.

Quantitative systems pharmacology modeling of tumor heterogeneity in response to BH3-mimetics using virtual tumors calibrated with cell viability assays

Both primary and acquired resistance mechanisms that involve intra-tumoral cell heterogeneity limit the use of BH3-mimetics to trigger tumor cell apoptosis. This article proposes a new quantitative systems pharmacology (QSP)-based methodology in which cell viability assays are used to calibrate virtual tumors (VTs) made of virtual cells whose fate is determined by simulations from an apoptosis QSP model. VTs representing SU-DHL-4 and KARPAS-422 cell lines were calibrated using in vitro data involving venetoclax (anti-BCL2), A-1155463 (anti-BCLXL), and/or A-1210477 (anti-MCL1). The calibrated VTs provide insights into the combination of several BH3-mimetics, such as the distinction between cells eliminated by at least one of the drugs (monotherapies) from the cells eliminated by a pharmacological combination only. Calibrated VTs can also be used as initial conditions in an agent-based model (ABM) framework, and a minimal ABM was developed to bridge in vitro SU-DHL-4 cell viability results to tumor growth inhibition experiments in mice.

Quantitative systems pharmacology modeling of macrophage-targeted therapy combined with PD-L1 inhibition in advanced NSCLC

Immune checkpoint inhibitors remained the standard-of-care treatment for advanced non-small cell lung cancer (NSCLC) for the past decade. In unselected patients, anti-PD-(L)1 monotherapy achieved an overall response rate of about 20%. In this analysis, we developed a pharmacokinetic and pharmacodynamic module for our previously calibrated quantitative systems pharmacology model (QSP) to simulate the effectiveness of macrophage-targeted therapies in combination with PD-L1 inhibition in advanced NSCLC. By conducting in silico clinical trials, the model confirmed that anti-CD47 treatment is not an optimal option of second- and later-line treatment for advanced NSCLC resistant to PD-(L)1 blockade. Furthermore, the model predicted that inhibition of macrophage recruitment, such as using CCR2 inhibitors, can potentially improve tumor size reduction when combined with anti-PD-(L)1 therapy, especially in patients who are likely to respond to anti-PD-(L)1 monotherapy and those with a high level of tumor-associated macrophages. Here, we demonstrate the application of the QSP platform on predicting the effectiveness of novel drug combinations involving immune checkpoint inhibitors based on preclinical or early-stage clinical trial data.

In Silico Clinical Trials: Is It Possible?

Modeling and simulation (M&S), including in silico (clinical) trials, helps accelerate drug research and development and reduce costs and have coined the term "model-informed drug development (MIDD)." Data-driven, inferential approaches are now becoming increasingly complemented by emerging complex physiologically and knowledge-based disease (and drug) models, but differ in setup, bottlenecks, data requirements, and applications (also reminiscent of the different scientific communities they arose from). At the same time, and within the MIDD landscape, regulators and drug developers start to embrace in silico trials as a potential tool to refine, reduce, and ultimately replace clinical trials. Effectively, silos between the historically distinct modeling approaches start to break down. Widespread adoption of in silico trials still needs more collaboration between different stakeholders and established precedence use cases in key applications, which is currently impeded by a shattered collection of tools and practices. In order to address these key challenges, efforts to establish best practice workflows need to be undertaken and new collaborative M&S tools devised, and an attempt to provide a coherent set of solutions is provided in this chapter. First, a dedicated workflow for in silico clinical trial (development) life cycle is provided, which takes up general ideas from the systems biology and quantitative systems pharmacology space and which implements specific steps toward regulatory qualification. Then, key characteristics of an in silico trial software platform implementation are given on the example of jinkō.ai (nova's end-to-end in silico clinical trial platform). Considering these enabling scientific and technological advances, future applications of in silico trials to refine, reduce, and replace clinical research are indicated, ranging from synthetic control strategies and digital twins, which overall shows promise to begin a new era of more efficient drug development.

Quantitative Systems Pharmacology and Machine Learning: A Match Made in Heaven or Hell?

As pharmaceutical development moves from early-stage in vitro experimentation to later in vivo and subsequent clinical trials, data and knowledge are acquired across multiple time and length scales, from the subcellular to whole patient cohort scale. Realizing the potential of this data for informing decision making in pharmaceutical development requires the individual and combined application of machine learning (ML) and mechanistic multiscale mathematical modeling approaches. Here we outline how these two approaches, both individually and in tandem, can be applied at different stages of the drug discovery and development pipeline to inform decision making compound development. The importance of discerning between knowledge and data are highlighted in informing the initial use of ML or mechanistic quantitative systems pharmacology (QSP) models. We discuss the application of sensitivity and structural identifiability analyses of QSP models in informing future experimental studies to which ML may be applied, as well as how ML approaches can be used to inform mechanistic model development. Relevant literature studies are highlighted and we close by discussing caveats regarding the application of each approach in an age of constant data acquisition. SIGNIFICANCE STATEMENT: We consider when best to apply machine learning (ML) and mechanistic quantitative systems pharmacology (QSP) approaches in the context of the drug discovery and development pipeline. We discuss the importance of prior knowledge and data available for the system of interest and how this informs the individual and combined application of ML and QSP approaches at each stage of the pipeline.

Informing virtual clinical trials of hepatocellular carcinoma with spatial multi-omics analysis of a human neoadjuvant immunotherapy clinical trial

Human clinical trials are important tools to advance novel systemic therapies improve treatment outcomes for cancer patients. The few durable treatment options have led to a critical need to advance new therapeutics in hepatocellular carcinoma (HCC). Recent human clinical trials have shown that new combination immunotherapeutic regimens provide unprecedented clinical response in a subset of patients. Computational methods that can simulate tumors from mathematical equations describing cellular and molecular interactions are emerging as promising tools to simulate the impact of therapy entirely in silico. To facilitate designing dosing regimen and identifying potential biomarkers, we developed a new computational model to track tumor progression at organ scale while reflecting the spatial heterogeneity in the tumor at tissue scale in HCC. This computational model is called a spatial quantitative systems pharmacology (spQSP) platform and it is also designed to simulate the effects of combination immunotherapy. We then validate the results from the spQSP system by leveraging real-world spatial multi-omics data from a neoadjuvant HCC clinical trial combining anti-PD-1 immunotherapy and a multitargeted tyrosine kinase inhibitor (TKI) cabozantinib. The model output is compared with spatial data from Imaging Mass Cytometry (IMC). Both IMC data and simulation results suggest closer proximity between CD8 T cell and macrophages among non-responders while the reverse trend was observed for responders. The analyses also imply wider dispersion of immune cells and less scattered cancer cells in responders' samples. We also compared the model output with Visium spatial transcriptomics analyses of samples from post-treatment tumor resections in the original clinical trial. Both spatial transcriptomic data and simulation results identify the role of spatial patterns of tumor vasculature and TGFβ in tumor and immune cell interactions. To our knowledge, this is the first spatial tumor model for virtual clinical trials at a molecular scale that is grounded in high-throughput spatial multi-omics data from a human clinical trial.

Using machine learning surrogate modeling for faster QSP VP cohort generation

Virtual patients (VPs) are widely used within quantitative systems pharmacology (QSP) modeling to explore the impact of variability and uncertainty on clinical responses. In one method of generating VPs, parameters are sampled randomly from a distribution, and possible VPs are accepted or rejected based on constraints on model output behavior. This approach works but can be inefficient (i.e., the vast majority of model runs typically do not result in valid VPs). Machine learning surrogate models offer an opportunity to improve the efficiency of VP creation significantly. In this approach, surrogate models are trained using the full QSP model and subsequently used to rapidly pre-screen for parameter combinations that result in feasible VPs. The overwhelming majority of parameter combinations pre-vetted using the surrogate models result in valid VPs when tested in the original QSP model. This tutorial presents this novel workflow and demonstrates how a surrogate model software application can be used to select and optimize the surrogate models in a case study. We then discuss the relative efficiency of the methods and scalability of the proposed method.

Generating immunogenomic data-guided virtual patients using a QSP model to predict response of advanced NSCLC to PD-L1 inhibition

Generating realistic virtual patients from a limited amount of patient data is one of the major challenges for quantitative systems pharmacology modeling in immuno-oncology. Quantitative systems pharmacology (QSP) is a mathematical modeling methodology that integrates mechanistic knowledge of biological systems to investigate dynamics in a whole system during disease progression and drug treatment. In the present analysis, we parameterized our previously published QSP model of the cancer-immunity cycle to non-small cell lung cancer (NSCLC) and generated a virtual patient cohort to predict clinical response to PD-L1 inhibition in NSCLC. The virtual patient generation was guided by immunogenomic data from iAtlas portal and population pharmacokinetic data of durvalumab, a PD-L1 inhibitor. With virtual patients generated following the immunogenomic data distribution, our model predicted a response rate of 18.6% (95% bootstrap confidence interval: 13.3-24.2%) and identified CD8/Treg ratio as a potential predictive biomarker in addition to PD-L1 expression and tumor mutational burden. We demonstrated that omics data served as a reliable resource for virtual patient generation techniques in immuno-oncology using QSP models.

Generating immunogenomic data-guided virtual patients using a QSP model to predict response of advanced NSCLC to PD-L1 inhibition

Generating realistic virtual patients from a limited amount of patient data is one of the major challenges for quantitative systems pharmacology modeling in immuno-oncology. Quantitative systems pharmacology (QSP) is a mathematical modeling methodology that integrates mechanistic knowledge of biological systems to investigate dynamics in a whole system during disease progression and drug treatment. In the present analysis, we parameterized our previously published QSP model of the cancer-immunity cycle to non-small cell lung cancer (NSCLC) and generated a virtual patient cohort to predict clinical response to PD-L1 inhibition in NSCLC. The virtual patient generation was guided by immunogenomic data from iAtlas portal and population pharmacokinetic data of durvalumab, a PD-L1 inhibitor. With virtual patients generated following the immunogenomic data distribution, our model predicted a response rate of 18.6% (95% bootstrap confidence interval: 13.3-24.2%) and identified CD8/Treg ratio as a potential predictive biomarker in addition to PD-L1 expression and tumor mutational burden. We demonstrated that omics data served as a reliable resource for virtual patient generation techniques in immuno-oncology using QSP models.

A quantitative systems pharmacology model of the pathophysiology and treatment of COVID-19 predicts optimal timing of pharmacological interventions

A quantitative systems pharmacology (QSP) model of the pathogenesis and treatment of SARS-CoV-2 infection can streamline and accelerate the development of novel medicines to treat COVID-19. Simulation of clinical trials allows in silico exploration of the uncertainties of clinical trial design and can rapidly inform their protocols. We previously published a preliminary model of the immune response to SARS-CoV-2 infection. To further our understanding of COVID-19 and treatment, we significantly updated the model by matching a curated dataset spanning viral load and immune responses in plasma and lung. We identified a population of parameter sets to generate heterogeneity in pathophysiology and treatment and tested this model against published reports from interventional SARS-CoV-2 targeting mAb and antiviral trials. Upon generation and selection of a virtual population, we match both the placebo and treated responses in viral load in these trials. We extended the model to predict the rate of hospitalization or death within a population. Via comparison of the in silico predictions with clinical data, we hypothesize that the immune response to virus is log-linear over a wide range of viral load. To validate this approach, we show the model matches a published subgroup analysis, sorted by baseline viral load, of patients treated with neutralizing Abs. By simulating intervention at different time points post infection, the model predicts efficacy is not sensitive to interventions within five days of symptom onset, but efficacy is dramatically reduced if more than five days pass post symptom onset prior to treatment.

Virtual clinical trials of anti-PD-1 and anti-CTLA-4 immunotherapy in advanced hepatocellular carcinoma using a quantitative systems pharmacology model

BACKGROUND

Hepatocellular carcinoma (HCC) is the most common form of primary liver cancer and is the third-leading cause of cancer-related death worldwide. Most patients with HCC are diagnosed at an advanced stage, and the median survival for patients with advanced HCC treated with modern systemic therapy is less than 2 years. This leaves the advanced stage patients with limited treatment options. Immune checkpoint inhibitors (ICIs) targeting programmed cell death protein 1 (PD-1) or its ligand, are widely used in the treatment of HCC and are associated with durable responses in a subset of patients. ICIs targeting cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) also have clinical activity in HCC. Combination therapy of nivolumab (anti-PD-1) and ipilimumab (anti-CTLA-4) is the first treatment option for HCC to be approved by Food and Drug Administration that targets more than one immune checkpoints.

METHODS

In this study, we used the framework of quantitative systems pharmacology (QSP) to perform a virtual clinical trial for nivolumab and ipilimumab in HCC patients. Our model incorporates detailed biological mechanisms of interactions of immune cells and cancer cells leading to antitumor response. To conduct virtual clinical trial, we generate virtual patient from a cohort of 5,000 proposed patients by extending recent algorithms from literature. The model was calibrated using the data of the clinical trial CheckMate 040 (ClinicalTrials.gov number, NCT01658878).

RESULTS

Retrospective analyses were performed for different immune checkpoint therapies as performed in CheckMate 040. Using machine learning approach, we predict the importance of potential biomarkers for immune blockade therapies.

CONCLUSIONS

This is the first QSP model for HCC with ICIs and the predictions are consistent with clinically observed outcomes. This study demonstrates that using a mechanistic understanding of the underlying pathophysiology, QSP models can facilitate patient selection and design clinical trials with improved success.