Research project, early stage

Digital twin of human physiology

One coupled system of equations in which the body's fast processes and its slow ones share a single state and shape each other, personalised by a person's own data, to answer "what happens if", not only "what happened".

This is a research ambition, not a product. We describe what we want to build, how we test each step, and where the limits of our knowledge are.

Why connect fast and slow processes

Most health data describes either the next few hours or the next decades, and the two rarely meet. Heart rate, blood pressure, sleep, heart rate variability and glucose change within hours and days. Atherosclerosis, thickening of the heart muscle and metabolic age change over years. The slow processes are built from the fast ones, yet they are usually measured and discussed apart.

Hours and days

  • Heart rate and blood pressure
  • Sleep and heart rate variability
  • Glucose
  • Physical load

Years

  • Atherosclerosis
  • Thickening of the left ventricle
  • Metabolic age

The goal is one coupled multiscale system of equations. Fast variables (heart rate, blood pressure, glucose, sleep, heart rate variability) and slow ones (the vessel wall, the mass of the heart muscle, metabolic age) share one state and influence each other in both directions. Blood pressure reshapes the heart muscle over time, and the reshaped muscle in turn changes blood pressure. A question like "what changes if blood pressure stays lower for a year" should then have an answer with an honest range around it.

How we approach it

The system is one, but it cannot be built or trusted all at once. We implement and check it block by block. A block is a group of equations within the same system, never a separate model: it reads and writes the shared state that the other blocks use.

  • Each block has a defined input: the measurements it needs and how often.
  • Each block states its hypothesis: how it moves the shared state, and how the rest of the state moves it.
  • Each block is checked forecast against actual: the system predicts, the person's later measurements arrive, and we record the error.
  • Uncertainty is always shown as a range, never as a single number.

The long-term research goal

In the long run the same personalisable system should span every level of the body: from molecules and cells to organs and the whole organism. The broader direction is our own drugs and new methods of treatment for priority diseases. The twin is the research instrument for that: it tests hypotheses about mechanisms and about how an individual might respond. Known drugs serve as one of the references for checking it, not as the end goal. This describes a goal: it is not drug development, not an experiment and not a medical intervention, and no product or treatment is offered.

Levels of one system

  1. Molecules
  2. Cells
  3. Organs
  4. The whole body

The research path

  • Pharmacokinetics and pharmacodynamics (PK/PD) and quantitative systems pharmacology (QSP) join as blocks of the same system, not as separate models.
  • The first step is to reproduce an effect that is already known, on published data, and to have it checked independently, with the uncertainty stated explicitly.
  • Nothing on this path is tested on people and nothing here suggests any treatment.

Blocks of one system

These are the blocks we start with. Each one is a part of the same system and shares its state with the others. The list will change as the checks show what holds up.

  1. Lipids and the vessel wall

    How long-term levels of ApoB and LDL cholesterol relate to the growth of plaque in the arteries. A first prototype of this block exists for internal research; it has no validated results.

  2. Blood pressure and the left ventricle

    How sustained blood pressure thickens the heart's main pumping chamber, and how a thicker chamber changes blood pressure in return. To be checked against repeated heart ultrasound.

  3. Sleep, variability, load and recovery

    How sleep, heart rate variability and physical load from wearable devices move day-to-day recovery, and through it the slower parts of the state.

  4. Glucose, insulin resistance and metabolic age

    How glucose, insulin and HbA1c move together, and how insulin resistance relates to markers of metabolic ageing.

  5. The shared state

    The variables all blocks read and write, and the rules that keep processes on different time scales consistent with each other.

Where the project stands

What exists

  • The research goal: one coupled system, and the way each of its blocks will be checked.
  • A first prototype of one block, used only for internal research.

What does not exist

  • A finished or validated system.
  • Any clinical or diagnostic use: the system does not diagnose and does not suggest treatment.
  • Regulatory approval of any kind.

Open questions

  • How much data a person needs before the system's forecast is better than a simple average.
  • How to couple processes that run on such different time scales without the errors adding up.
  • How well a system tuned on one group of people carries over to another.
  • Which measurements are worth their cost and which add noise.

How we measure progress

  • For each block and for the system as a whole: the error of the forecast against what was actually measured later, on real series of data.
  • For each block: the share of people who have enough data for it to work at all.
  • The width of the uncertainty range, which should narrow as data accumulates.

Open physiology models as a reference

Open physiology engines, such as Pulse, are a reference point for this research: they show how the body's systems can be described in equations. We do not depend on any one of them and can replace or combine them as the work requires.

This is a research project. It is not medical advice, a diagnosis or a treatment recommendation. Decisions about your health belong with you and your doctor.

Digital twin of human physiology — research project | Lonevi