Main text

Airway diseases account for one of the highest causes of mortality and morbidity worldwide1. Lung diseases may arise congenitally from genetic mutations or be acquired through acute exposures and environmental triggers. In 2019, chronic obstructive pulmonary disease (COPD), lung cancer, and acute lower respiratory infections collectively accounted for over 8 million deaths worldwide, ranking among the leading causes of mortality2. The global increase of respiratory diseases from 2019 to 2021 is mirrored by an increase in global healthcare spending, which grew from 8.5 trillion USD (9.8% global GDP) to 9.8 trillion USD (10.3% global GDP)3. In clinical settings, disease management strategies range from symptom-managing medications to intensive care3,4. Severe cases often necessitate surgical intervention, such as mechanical ventilation and, in late-stage diseases, lung transplantation when available. The growing prevalence and burden of lung-related diseases underscores the urgent need for advanced predictive modeling tools.

A significant challenge in developing effective modeling tools lies in the limited understanding of fundamental biological pathways and their mechanisms within the lung. Many of these pathways are disrupted in disease or activated as part of the body’s response to it, necessitating an in-depth investigation into the complex cellular interactions, regulatory networks, and their downstream effects5,6. Consequently, researchers have increasingly focused on studying lung development, where dynamic biological pathways drive cellular differentiation and regulated proliferation. This approach has provided valuable insights into disease mechanisms, such as the proliferative behavior of cancer cells and the effects of specific mutations in congenital disorders7.

Animal models are widely used to study disease pathology and test potential therapeutic strategies; however, human-specific differences in disease manifestation often limit their relevance to humans. A meta-analysis of over 100 systematic reviews looking at the translational potential of animal studies to human treatments found that only 5% of interventions validated in animal models were successfully translated to approved human medicines8. While human-based models, such as primary cells and tissues, address some of the limitations, their use has been constrained by availability, especially in rare diseases9. Human cell lines are more accessible and commonly used, but their lack of tissue-level complexity restricts their utility10. This underscores the need for improved models that better capture the biological and functional complexities of the lung or its specific components.

Human pluripotent stem cells (hPSCs), offer a promising solution to generate tissue-level models for research. These stem cells are capable of self-renewal and differentiation and offer a great resource to study human lung development, model respiratory diseases, and identification and testing of novel therapeutic interventions11,12,13,14,15. In generating hPSC-derived lung models (especially more complex models such as “organoids”), the steps follow a directed developmental trajectory, which enables the generation of key cell states such as tissue-specific progenitor cells with fetal characteristics. While for cell regeneration, maturation of these hPSC-derived cells has largely been an issue, an advantage of fetal-like cells is the ability to study developmental processes (i.e., pathways that drive specific cell lineage differentiation and maturation). Moreover, access to primary fetal tissues is restricted and limited16,17,18. With the nobel-winning discovery of induced pluripotent stem cells (hiPSC), generation of fetal-like cells to model congenital disorders caused by genetic mutations is now possible and will advance new research to identify therapies especially for rare or ultra-rare diseases19,20,21. Induced pluripotent stem cells are artificially made through cellular reprogramming and are a powerful resource for future precision medicine. An example is the use of hiPSC derived from patients with a life-limiting genetic disease called cystic fibrosis (CF) in testing new small molecule therapies22,23. CF is a disease whereby mutations in the gene encoding a chloride channel CFTR (cystic fibrosis transmembrane conductance regulator) lead to defective ion and water transport across the airway epithelium. This results in mucous accumulation and plugging fostering persistent respiratory infections, and chronic inflammation. A major advantage of hiPSC-derived lung models is the ability to develop high-throughput platforms to test novel therapies.

Recent high-throughput next-generation sequencing technologies (omics) have made it possible to analyze the transcriptomic and epigenomic states of cells at single-cell resolution, even with limited samples. These advancements have enabled the generation of extensive datasets that researchers can use to create mathematical and computational models of cellular states during homeostasis, injury, and disease24. Building on these developments, researchers are working toward creating a holistic predictive model, known as a digital twin. This model aims to faithfully replicate biological contexts, including structural anatomy, cellular behavior, and organ functionality, in silico25. When accurately constructed, the digital twin becomes a powerful tool for studying perturbations associated with genetic or acquired diseases, including respiratory conditions. A key advantage of the digital twin lies in its ability to predict treatment outcomes and guide clinical decisions in a personalized manner26. This is particularly valuable for studying rare biological phenomena, such as congenital diseases, and for addressing urgent needs in acute conditions, such as during the COVID-19 pandemic.

To date, several initiatives have been established to generate and process single-cell datasets for the creation of digital twins. One such initiative is the Chan Zuckerberg Initiative (CZI) Cell Science program, which aims to foster interdisciplinary collaborations to deepen our understanding of cellular processes and disease pathology. The program focuses on collecting comprehensive cellular data, engineering predictive models, and applying them to various disease models27. Notable lung-related projects supported by this program include the creation of atlases, such as the “Inflammatory Childhood Interstitial Lung Disease Cell Atlas,” and disease-focused studies using primary human and animal samples, including “A Data-Driven Approach to Understanding Liver and Lung Diseases” and “Understanding How Age Influences Susceptibility to Coronavirus Infection.” Similarly, the European Virtual Human Twin (VHT) initiative, led by EDITH, aims to create a digital twin for patient-specific predictive modeling, with a focus on diseases like cancer, cardiovascular conditions, intensive care, osteoporosis, and brain disorders28. VHT integrates clinical data through smart sensors, imaging, and nanomedicine, alongside omics and in silico data, to develop comprehensive models. While these initiatives differ in focus, both recognize the importance of a multidisciplinary approach. Progress in creating accurate digital twins can only be achieved by combining a deep understanding of biology with advanced computational and mathematical modeling. In this context, we outline our current understanding of lung biology, in vitro models, and in silico models that are essential for advancing the creation of a digital lung twin.

Basic structure of the lung and its cell types

The anatomical structure of the human lung is crucial for its role in gas exchange29. Along the proximal-distal axis, the proximal conducting region humidifies air and removes pathogens, while the distal respiratory region facilitates gas exchange between the external environment and the bloodstream30. In the alveoli, oxygen from the inhaled air is exchanged for carbon dioxide from the blood (Fig. 1). The differing roles within the proximal-distal regions of the lung is further reflected in its differences in cell type composition and organization. From the proximal to distal direction, the lung transitions from a thick pseudostratified epithelium, simple columnar epithelium, and thin alveolar sacs. Proper lung function relies on the interaction of diverse cell types, making their study essential for developing models of lung development and disease.

Fig. 1: Anatomical and cellular composition of the human lung, highlighting its role in gas exchange and defense.
figure 1

The lung is structured into proximal and distal regions. The proximal airway (top right) humidifies air and removes pathogens, supported by epithelial cells that clear foreign matter and transmit environmental signals, mesenchymal cells providing structural support, endothelial cells aiding in gas exchange, and immune cells defending against pathogens. The distal airway (bottom right) facilitates gas exchange through specialized epithelial cells and stromal support. The alveoli (bottom left) are the site of oxygen and carbon dioxide exchange between inhaled air and pulmonary capillaries, with collaboration between epithelial cells (AT1 and AT2) and endothelial cells.

Full size image

Lung cell populations can be broadly categorized into four groups: Epithelial cells that line the airway surface and is the first line of defense against inhaled pathogens, support gas exchange, clear foreign matter and transmit environmental signals to the brain and circulatory system (Fig. 1). Mesenchymal cells contribute to the lung development by providing differentiation cues and nutrients31. Endothelial cells, which line the pulmonary capillaries, collaborate with epithelial cells in gas exchange. Immune cells defend against pathogens in inhaled air. Each category contains specialized subpopulations with distinct roles and molecular characteristics.

The roles of epithelial cells vary by their location in the lung. Here, we briefly describe some of the major cell types in specific regions of the airway. Description of additional cell types and their localization and function can be found in detailed reviews32,33. The airway epithelium conducts air into the lungs and protects against pathogens34. Major epithelial cell types that line the airway epithelium include basal cells, a subset of which are resident stem cells along the basolateral epithelial layer protected from the external environment. These cells respond to injury and undergo differentiation to regenerate the airway epithelium35. Club cells are a type of secretory cells that secrete cytochrome P450 enzymes and other proteases and support metabolism, detoxification and immune protection of the airways36. Ciliated cells are involved in the physical removal of inhaled pathogens through mucociliary transport37. Decrease in club and ciliated cell populations have been implicated in diseases like COPD and CF38,39. Goblet cells are another type of secretory cells that produce mucus and have been shown to contribute to the maintenance of ciliated cells under normal conditions40.

The alveolar epithelium is responsible for gas exchange. There are two types of epithelial cells in the alveoli: Type I pneumocytes (AT1) and Type II pneumocytes (AT2) (Fig. 1). The incredibly thin morphology of the AT1 cells enables gas exchange with closely associated capillaries and oxygenating the blood within them. These cells make up much of the alveolar epithelium and account for approximately 95% of the epithelial cells in this region30. On the contrary, AT2 cells make up a smaller proportion of the alveolar epithelium but play an important role in the biomechanics of breathing as these cells produce surfactant proteins required to regulate surface tension to permit lung expansion and proteins to regulate immune cell function41. These cells also serve as resident progenitor cells that can give rise to AT1 cells in the context of repair30. Both of these cell types have been implicated in diseases such as COPD, neonatal respiratory distress syndrome caused by surfactant insufficiency and respiratory infections41.

Mesenchymal cells encompass a large population of different stromal cell types such as smooth muscle cells, airway fibroblasts, myofibroblasts. New single cell technologies are only beginning to unravel all the different mesenchymal cell subtypes that exist and their specific functional role during development and injury/repair has yet to be clearly defined. Mesenchymal cells play an important role in lung development by supplying epithelial cells with essential soluble (ie. growth factors) and structural (ie. extracellular matrices) cues and nutrients to support differentiation and tissue growth31. Detailed characterization of some of the mesenchymal cell types are reviewed in ‘Mesenchymal Cells in The Lung: Evolving Concepts and Their Role in Fibrosis’42.

Endothelial cells line the capillaries in the pulmonary system, closely associating with AT1 cells to mediate gas exchange. In large blood vessels that associate with the bronchioles, these cells support immune regulation by recruiting and enabling passage of circulating immune cells43. Dysregulation of endothelial function can impact homeostasis between lung compartments and blood, which can lead to Acute Respiratory Distress Syndrome (ARDS) and pulmonary hypertension43. ARDS is the result of alveoli filling with fluid as a result of inflammation in the endothelium44.

Immune cells protect the lungs from inhaled pathogens, maintaining respiratory health. During development, these immune cells that exist in the developing fetal lung circulation migrate to the developing lung parenchyma and reside there until birth. The lung is home to some of the key immune cell types, including alveolar macrophages, leukocytes and neutrophils41. Other than immune regulation, recent studies have identified developmental roles of IL-1β-secreting myeloid cells in basal cell differentiation45, suggesting the lung immune cells that co-develop during development have pleiotropic roles in formation of the airways, maintaining lung homeostasis as well as defense.

In summary, by identifying all the cells that exist in the human lung and investigating the gene regulatory networks that drive their development using relevant experimental models, researchers can better simulate how these cells contribute to development, homeostasis and disease, and in turn, pave the way for improved diagnostic or therapy discovery. With new technologies and improved models, researchers will continue to uncover deep insights into the lung’s complexity. As our understanding of the lung’s cellular landscape evolves, there is a need for modeling efforts to include these complexities which can be achieved with close collaborations between experimentalists and modeling scientists.

Human in vitro lung models

Much of the current knowledge of lung development is derived from mouse models. While there are many similarities in cell types between humans and mice, significant differences exist in lung morphology46. For instance, human lungs have more cartilage, additional airway branches, and greater capacity than mouse lungs, presenting challenges that mouse models cannot fully address. Therefore, there is a pressing need to develop biological models that better replicate the human lung to study development and disease.

When creating a biological model, it is important to consider factors such as scale, anatomical accuracy, and the necessary cell populations, depending on which section of the lung is being modeled47. There are a range of lung cell models that are currently used and can generally fall into two categories: two-dimensional (2D) cell cultures and three-dimensional (3D) models. Traditional 2D lung models culture a homogenous population of cells in a submerged monolayer format and have been useful in studying basic functions of lung cells and differentiation processes48. To further the modeling complexity and its representation of the airway microenvironment, lung cells are cultured in an air-liquid interface (ALI) format where the epithelial cells are exposed to air on the apical membrane and culture media on the basolateral side. This format is commonly used to induce epithelial polarization and formation of complex epithelia, such as the pseudostratified epithelium of the large airways. The ALI model has been used to assess air surface liquid production, ion channel function, and ciliary movement49,50,51. On the other hand, 3D lung models such as “organoids” leverage the self-organizing potential of stem cells to form enclosed 3D structures. Simple epithelial organoids often form spheroids and when co-cultured with lung stromal cells, form tube-like structures reminiscent of airway branches. This technique can be used to study branching morphogenesis (formation of de novo airways)52. In both models, the use of stem cells enables the generation of multiple lung cell types that co-develop and can be used to study cell autonomous (ie. cell fitness) and non-cell autonomous effects (i.e., cell competition). Both 2D and 3D models have been used for disease modeling and therapy discovery for lung diseases such as CF and respiratory infections13,14,23,50,53.

Another in vitro model that has been traditionally used for lung research is the lung explant model. Explants are thin slices of fresh primary tissues made using a tissue slicer and are called precision cut lung slices. Depending on how the lungs explants are cut, they retain most, if not all of the cells in the native lung thereby leaving the cellular composition and tissue organization intact. As a result, explants have been used to study airway branching morphogenesis54,55. While lung explants are powerful models, they can only be cultured ex-vivo for a limited time and therefore preclude their use in long-term studies54.

Biological in vitro models of the lung are invaluable for studying development and disease; however, several limitations must be acknowledged to improve current models. While 2D models are useful, they lack important cellular interactions and biophysical cues which limit their in vivo applications56. 3D organoid models have improved upon 2D models through introducing structural complexity and cellular heterogeneity; however, these models lack several key lung components, including endothelial and immune cells, vascular networks, and the diverse ECM proteins that support the lungs57,58. As a result, these models may not fully recapitulate lung physiology or accurately predict disease states, particularly in immune-related diseases where the immune system plays a critical role in disease progression. Furthermore, 3D models are constrained in size, as they rely on diffusion to deliver nutrients, limiting their ability to model the challenges posed by the lung’s large volume56. It is necessary to acknowledge the current limitations of these models when exploring multi-scale lung modeling and leverage emerging biological models to iteratively refine the predictive model and ensure accurate representations of the human lung biology.

Despite the limitations in in vitro modeling, the models have been pivotal in progressing research in developmental lung biology. The generation of powerful in vitro lung models has been greatly informed by findings in developmental lung biology. Insights from studying developmental biological processes have been critical in optimizing cell media compositions and engineering matrices that support cell growth59,60. This is particularly true in hPSC-derived models where many signaling pathways identified to be important during development are leveraged to direct efficient differentiation of lung cells16,61,62. These directed differentiations of hPSC offer an unprecedented resource to study new pathways that drive specific cell fates and as such can be used to generate targeted cell types for future tissue regeneration. Indeed, researchers are using hPSC-derived fetal lung progenitors to study complex processes such as branching morphogenesis and cellular differentiations63,64 by identifying the key signaling pathways that drive lung fate during lung organogenesis65.

An emerging in vitro lung platform that builds upon the 2D and 3D models by introducing dynamic flow are called lung-on-chip microfluidic devices58. Bioprinting combined with “on-chip” designs are advancing “3D” models to including physiological stressors that are important during development and impact disease66,67. While exact designs can vary, cells are housed within a device that has channels that connect to the environment in a circuit68. As a result, the cell’s microenvironment can be adjusted by pumping the channels with liquids or gases: used for mimicking blood flow or fetal breathing movements and used for recapitulating dynamic airflow in breathing respectively69,70,71. Advancements in microfluidic technologies have also introduced changes that recapitulate the lung architecture and physiological function through incorporating the tubular structure of airways and cell stretching during breathing66,72.

Advancements in in vitro lung platforms can be significantly accelerated by integrating in silico computational and mathematical modeling, which can help pinpoint specific pathways for experimentalists to focus on. In the following section of this perspective paper, we examine some of the existing in silico lung models, with the aim of combining in vitro and in silico approaches to develop a new generation of lung models that more accurately represent the human lung.

Human in silico lung models

The efficacy of a digital twin depends on its ability to properly capture and represent biological phenomena through mathematical and computational models. Currently, computational modeling of the lungs integrates various fields (e.g., biophysics, biomechanics, physiology) with large data (e.g., single cell genomics, medical imaging)73. By incorporating advanced algorithms and simulations, researchers can create representations of lung dynamics and function. Additionally, integrating medical imaging data, such as CT scans and MRIs, allows researchers to incorporate lung structures with more accuracy. As a result, computational modeling holds the potential to facilitate experimental strategies, optimize our discoveries, and ultimately lead to better clinical outcomes for individuals affected by lung diseases.

Various models are used to understand the mechanical properties and structures of the lung, such as spring models, micromechanical models and liquid film models73. Mathematical modeling of the mechanical forces in the lungs is intricate, and an effective computational simulation would depend on understanding the mechano-structural complexities at various scales. Computational models have been developed to simulate cell strain in different cell types, revealing that cell strain during fetal breathing and post-birth affects the differentiation of AT1 and AT2 cells74. For instance, Massida et al. created a computational model to predict the impact of mechanical waveforms, mimicking physiological conditions, on chromatin structure75. The changes in chromatin structure are known to activate Yap/Taz signaling within the nucleus, which is associated with stem cell differentiation76,77. When modeling the lung parenchyma, a computational approach based on a 3D Voronoi diagram was used to simulate stretching forces acting on the tissue at both the micro and macro scales78. This model demonstrated that the position of the cells within the alveolus significantly influenced the mechanical forces they experienced78. Much work has also been done to model lung function at the macro-scale, particularly in relation to airflow, drug-particle deposition, and mechanical damage79,80. Using computed tomography images of the upper and lower respiratory tract, researchers constructed a computational mesh to model the airflow and predict the velocity distribution, particle trajectories, and deposition zones within the respiratory system80. By adapting the computational mesh iteratively, researchers ensured accuracy in capturing airflow dynamics while reducing computational complexity80. This approach allows for detailed analysis of airflow patterns and drug-particle interactions, contributing to the development of effective drug delivery systems.

While mechano-properties are important for lung development, the underlying tissue and cellular complexities extend across many other biological processes (e.g., branching, gene interactions, cell interactions, and surfactant production)81,82. These complex biological processes have been successfully modeled computationally and mathematically by using morphogen gradients, gene regulation, and signal transduction83,84. For example, branching morphogenesis of the lung is intricately regulated by genetic interactions that produce signals resulting in the observed patterns81. Mathematical concepts such as fractal capture the architecture and the various types of bifurcations (orthogonal versus planar) while a FGF10 gene signaling pathway directs how cells grow81. This interplay across scales is key to understanding any biological phenomenon and drives the current initiatives to construct accurate gene regulatory networks (GRNs). In GRNs, genes are represented as nodes, and their interactions are shown as edges. These networks can also integrate transcription factors, splicing factors, non-coding RNAs, microRNAs, and metabolites84. With the growing availability of scRNA-seq data, several methods have been developed to infer GRNs in various biological systems85,86, offering deeper insights into gene regulation compared to relying solely on gene interaction databases84. In developmental contexts, GRNs have shown significant predictive power. For example, in pancreatic development, GRNs have enhanced our understanding of the development of Type 1 Diabetes87. In lung development, the impact of specific genes and transcription factors has been experimentally explored, providing an initial framework for understanding GRNs in the lung82. These gene interactions are foundational, but to fully leverage the predictive potential of GRNs in lung development, it is essential to incorporate the effects of structural and mechanical forces on this process.

Individually, in silico models tend to be limited in scope and are optimized for predicting specific outcomes. To overcome this, A digital twin that can encompass and predict complete lung function will require a collection of models, each excelling at predicting different aspects of development and clinical outcomes. Successful integration of these models will depend on identifying the interfaces between them. One coupled computational model aims to simulate pulmonary airflow and tissue deformation in the trachea, larger bronchi, terminal bronchioles, respiratory bronchioles, and alveoli88. To model the trachea and larger bronchi, a computational fluid dynamics model was created using the Navier-Stoke equation to simulate incompressible viscous flow with Reynolds-averaged turbulence modeling. For smaller airways a 1D pipe model was developed based on Poiseuille’ law to simulate airflow. In the respiratory bronchioles and alveoli, a biphasic mixture model was designed to simulate lung tissue deformation in response to airflow velocity and pressure. These models were coupled by transferring flow rates at larger bronchiole outlets to the inlets of the terminal bronchioles and using average pressures at larger bronchiole outlets as boundary conditions for the 1D pipe model. The 1D pipe model and biphasic mixture model were further linked by ensuring that the volume expansion of lung parenchyma matched the airflow from terminal bronchioles88. This coupled computational model is a strong example of integrating smaller models into a large computational framework. The coupling of these distinct models was made easier by simplifying assumptions, such as using a quasi-static state to model the distal lung regions, which reduces computational complexity but does not necessarily reflect in vivo conditions. Furthermore, pinpointing the precise junctions between the three models in the lung adds further complexity to its applicability. Overall, this model provides only a limited insight into the potential of multiscale models to simulate lung function on a larger scale.

In their review ‘Multi-scale lung modeling’, Tawhai and Bates highlight two crucial aspects in constructing effective multiscale models89. First, it is essential to identify the key properties of a complex system that emerge from the collective behavior of its components at a given scale, which must be integrated into higher levels. Second, it is important to determine which lower-level properties can be simplified or omitted without compromising the model’s accuracy89. For a successful digital twin of the developing human lung, it is vital to correctly identify the parameters where these multi-scale models interface. For example, how do macro-scale models of breathing mechanics relate to the micro-scale models focused on cell stretch and strain? How do these cellular mechanical changes affect chromatin structure and transcriptional function? A multi-scale model must also illustrate how a common phenomenon observed across several micro-scale models influences the macro-scale, and which of these phenomena can be neglected. For example, investigating how a non-functional cystic fibrosis transmembrane conductance regulator (CFTR) protein leads to mucus accumulation in respiratory bronchioles, which disrupts airflow, increases resistance, and ultimately impairs lung function at the macro-scale. Such a model would need to connect the molecular dysfunction of CFTR with the cascading effects on tissue mechanics, airway clearance, and gas exchange efficiency, showing the relationship between micro-scale biological disruption and their macro-scale physiological consequences (Fig. 2).

Fig. 2: Integration of biological and computational frameworks at three levels of scale can model the healthy human lung and generate a lung digital twin.
figure 2

At the system-level Navier-Stoke fluid Dynamic models can be used to model airflow and pressure within the larger airways. At the regional level, a 1D Pipe model can be used to model airflow based on Poiseuille’s law within the respiratory alveoli. Within the terminal bronchioles and alveoli, a Biphasic Mixture model can be used to simulate lung tissue deformation as a function of airflow and pressure. At the cellular-level, models can be used to predict the impact of pressure on alveolar stretching. At the subcellular-level, models capable of predicting the impact of cell stretching on gene expression can be utilized to inform GRNs within the cell.

Full size image

Looking ahead, mathematical and computational modeling of lung development offers significant potential for advancing our understanding of this complex process. Future efforts can harness recent advancements in multi-omics, spatial data, and live-cell imaging to create more comprehensive, data-driven models that integrate various biological layers. In silico models are also being developed for other organs; for example, a machine learning model has been used to predict heart disease with 80% accuracy90. Advances in heart in silico models can inform the improvement of lung in silico models, and vice versa. While experimental studies often guide and refine computational approaches, in silico methods have the power to generate hypotheses that can inform future experiments. These models will benefit from collaborative frameworks that combine iterative in vitro modeling with in silico modeling to foster a synergistic relationship where computational analysis informs experimental approaches to improve the efficacy and scope of a virtual lung. (Fig. 3). Moreover, a strong emphasis on validating and interpreting simulation results through rigorous experimental verification will ensure the reliability and relevance of these models to real-world biological contexts.

Fig. 3: Collaborative frameworks integrating in vitro modeling and in silico modeling to enhance the continuous collaborative development of a virtual lung.
figure 3

This synergy allows computational analysis to guide experimental approaches, improving efficacy and expanding the scope of lung modeling.

Full size image

Virtual Human Development Consortium

Recent advancements in technology have opened up new avenues for research. The emergence of sophisticated computational models allows researchers to simulate and analyze lung development in a virtual environment, providing valuable insights into the biological mechanisms at play. Producing a state-of-the-art computer-based simulator of human embryonic development is the mission of the Virtual Human Development (VHD) consortium (www.virtualhumandevelopment.org). The VHD is a new and growing consortium that aims to bridge various disciplines across a global community of scholars. By combining innovative approaches, it provides opportunities for researchers to push boundaries and shift in our understanding of developmental processes such as pulmonary architecture and function (Fig. 4). Researchers will leverage new and emerging technologies to investigate how the lungs develop and how perturbations in lung development or repair led to disease and impact therapies. These studies will help in the development of novel, more effective or targeted therapeutic interventions for respiratory diseases. The subgroup of VHD-Lung aims to predict cellular, mechanical and physiological functions of the lung at multiple resolutions. In silico models will utilize rules and algorithms derived from experimental models and the iterative process of modeling the lung as new data or advanced technologies emerge, will create a state-of-the-art virtual lung with the capability of determining a priori effects of disruptions/perturbations at different stages of human development.

Fig. 4: Combining multidisciplinary approaches across many different scales is integral to the construction of the VHD-lung.
figure 4

Cells represent the combined effort from multidisciplinary teams on multiple levels of scale to achieve a virtual human lung.

Full size image

The models and methods discussed in this perspective are not unique to the lung, as similar efforts have been made to develop virtual models for other major organs. For example, significant progress has been made in creating virtual twins of the pancreas, with applications in both diabetes and cancer research. With the increasing use of wearable technologies for diabetes management, a virtual pancreas twin has been developed to model complex blood glucose and insulin dynamics, enabling personalized lifestyle recommendations91,92. In pancreatic adenocarcinoma, the Molecular Twin leverages multi-omic data to identify critical biomarkers that predict patient outcomes93. In cardiac remodeling, researchers are integrating electrophysiological models with the biomechanical properties of the different heart chambers, to develop fully coupled models of the cardiovascular system.

Ultimately, the true potential of a virtual twin lies in its ability to link organ-level models and simulate the interconnected functions of all organs and tissues within the human body. For example, accurately modeling of the blood circulation within the cardiovascular system must be seamlessly connected to gas-exchange models of the respiratory system and the detoxification in the liver. Therefore, the greatest advantage and goal of a virtual human twin is to provide a holistic framework to study complex physiological processes.