Framework for personalised physiome modelling

The 12 LABOURS project develops and tests an energy-based, multiscale modelling framework that models the physiology—and therefore pathology—of any aspect of the human body. The framework incorporates person-specific data, including from wearable and implantable devices.

The Exemplar Projects (EPs) test this framework for selected, clinically relevant conditions using data from multiple spatial scales (blood biomarkers, clinical imaging of organ function, physiological function tests, and more). The project advances the long-standing Physiome Project, led for decades by the ABI, into a new, more holistic phase that aligns with the concept of a virtual human twin. Technology platform P1 develops this framework.

At the start of 12 LABOURS, we already have a well-developed modelling framework using CellML, the Physiome Model Repository (PMR), and associated tools such as OpenCOR. We also have many years of experience building spatial organ models that use partial differential equations to describe continuum physics at the organ scale.

Our organ-level models are designed to work with clinical imaging (MRI, CT etc.) to analyse tissue function, but they do not yet combine these data with blood biomarkers or genetic information—key molecular signatures that vary across individuals.

We also recognise that many published models (including those made reproducible and available in PMR) do not obey conservation laws of physics and therefore have limited use within an energy-based framework.

We realised that a simplified, physics-based approach can make subcellular modelling far more algorithmic. This results in a streamlined workflow for creating generic cell models and, subsequently, Functional Tissue Units (FTUs).

We begin by defining units suitable for all areas of physics relevant to physiological modelling. A port-Hamiltonian framework—built on these units—ensures conservation of energy, mass, and charge. The zero-dimensional form of this framework is the bond-graph (BG) approach. Several tools now handle both lumped-parameter BG models and higher-dimensional port-Hamiltonian models.

P1’s workflow proceeds through six key steps:

1. Six fundamental units define all physical processes within a mass-, charge-, and energy-conserving mathematical framework.
2. Each class of protein is captured by a single bond-graph template, with both a full kinetic and a reduced version.
3. Functional Cell Units (FCUs) assemble multiple protein-level bond-graph models. FCUs combine into Primary Functional Tissue Units (pFTUs), and sometimes secondary FTUs (sFTUs) at a higher spatial scale.
4. Organ models are built from FTUs. Some define constitutive laws for continuum models (e.g. finite-element PDEs), while others (e.g. the kidney) use geometric scaling from FTUs to whole-organ function.
5. Whole-body models integrate organ systems directly or via AI-fitted surrogates, using bond-graph systems-physiology models to represent regulatory mechanisms.

We define classes of proteins that require only one BG model to represent all members of that class. The protein classes shown in Figure 2 now have established templates.

For instance, one generic enzyme-catalysis model supports all metabolic pathways; one BG model represents the 485 proteins in the SLC transporter family; and one model captures all ion-channel mechanisms.

CellDL, an open-source tool by Dr Dave Brooks, builds these bond-graph protein models and exports them in CellML format for use with OpenCOR (Dr Alan Garny).

Roughly 2 % of the 3.2 billion base pairs in the human genome encode ~20 000 proteins. Another ~20 % encode RNAs that regulate transcription. Modelling these regulatory systems—capturing the physics of how RNAs control parameters of protein-level BG models—is crucial for achieving true multiscale integration.

FCUs represent subcellular functions by combining protein models into minimal functional modules.

Examples include:

• Muscle contraction FCU: models the actin–myosin system and calcium control (used in cardiac, skeletal, and smooth muscle).
• Metabolic FCU: models glycolysis, the TCA cycle, and the electron-transport chain using a single enzyme-catalysis template. This model runs > 1000 × faster than real time in OpenCOR on a single laptop core.

• Kidney FCUs: model reabsorption and excretion processes in the proximal tubule.

We are also building FCUs and FTUs for gas exchange (alveoli), vascular and airway smooth muscle (lungs), and lymphatic flow control.Although the kidney is not one of the 12 LABOURS EPs, it plays a key role in the Horizon Europe VITAL project, which tests scaling from FTU to organ level.

The Functional Tissue Unit (FTU) scale introduces diffusion constraints and the need for anatomical accuracy.

ABI’s software team develops the Physiological CAD tools (known as ABI Mapping Tools) that allow users to create anatomically detailed scaffolds at any scale, from FTUs to the whole body.

Scaffolds are constructed from geometric primitives (cubes, cylinders, spheres etc.) to form realistic organ shapes.

These scaffolds use high-order finite elements (cubic-Hermite basis functions) and can export computational meshes at any resolution or element type.

FTUs bridge the gap from cellular processes to organ-scale physiology, ensuring conservation of mass, charge, and energy.

We build nephron-level FTUs that combine reabsorption and excretion processes, including regulation via ADH and sympathetic control.

We also develop a generic, algorithmic method for FTU generation. The browser-based FTU Weaver tool converts bond-graph and port-Hamiltonian FCUs into composite port-Hamiltonian models (both graphically and via Python API). These tools [INSERT LINK HERE] are publicly available and engage the wider modelling community.

A workshop at ABI demonstrates FTU modelling with these tools. An exemplar cardiac electrophysiology model validates the framework.

We are now developing model-order-reduction methods to simplify FTU equations and to derive constitutive laws from composite port-Hamiltonians. These become part of the FTU workflow and will be released via the browser interface.

Cross-platform tools are also in development to ensure consistent user experience and exchange formats.

We identify key parameters in cardiac tissue (duty ratio, troponin concentration, Ca²⁺ sensitivity etc.) that must be preserved in FTU dynamics, and extend these workflows across other 12 LABOURS Exemplar Projects.

We develop a Functional Connectivity (FC) map (Figure 8) through the NIH-funded SPARC grant. This map serves as an interactive interface for systems-physiology models across all scales.

An example bond-graph systems-physiology model, integrating transport mechanisms in the kidney, liver, and gut, appears in Figure 9. We are implementing interactive visualisation so users can view simulation results in context with anatomical structures.

Finally, 3D anatomical models tailored to individuals are under development, drawing from the same knowledgebase used by the FC map.