ALTDATA Digital Twin · for biomedical research
Stop fighting the data gap. Start simulating it.
Physics-anchored AI that turns months of wet-lab iteration into virtual experiments you run in seconds, with a confidence score on every result.
The data gap
Your bottleneck isn’t vision. It’s data.
If you’re running a biotech program, the thing slowing you down isn’t ideas. It’s data. High-quality biological datasets are scarce. Wet-lab experiments take months, cost a fortune, and one slightly off variable can render an entire batch useless.
The waste is measurable, and it starts upstream. 90% of drugs that clear preclinical testing still fail in the clinic, and the correlation between in vivo results and what actually happens in trials runs under 8% (van Rijt et al., Cancers, 2023). A large part of that traces to the model itself: flat 2D cultures feed a uniform dose across a cell monolayer, nothing like the way a drug diffuses through real tissue, so they don’t predict physiological response with any accuracy. The model was wrong before the money was spent.
We don’t claim to move those numbers. We work on one piece of it: making the 3-D model you test against behave more like the biology, so fewer decisions rest on a result that was never going to hold.
That leaves most teams with two bad choices:
- Brute force. Burn your limited R&D budget on endless physical iterations, hoping you land the right configuration by chance.
- Black box. Lean on standard AI that looks great in a paper and fails in the lab, because it was never grounded in physics.
There’s a third way: the virtual experiment.
What it is, and what it isn’t
A validated engine, not a prediction toy.
The ALTDATA Digital Twin turns scarce biological data into scalable virtual experiments. It isn’t a prediction tool that guesses from patterns it has seen before. It’s anchored in physics: a high-fidelity physics simulator teaches a high-speed AI surrogate, so you get the rigour of a physics simulation at the speed of AI.
The practical difference: most AI twins are pattern recognisers. Ours is gated by physics, so it can’t hand you a result that’s physically impossible.
Scope today: the engine is validated for 3-D biological growth and structure over time, on one validated cell line. Broader biological systems are the platform roadmap, and we’ll tell you which is which.
Stop guessing
Test thousands of permutations in seconds, before you touch a single pipette.
Lower the cost of failure
Kill dead-ends virtually, so your wet-lab budget is spent only on high-probability winners.
Decide with confidence
Every result carries a confidence score, so you know when the model is certain and when it’s time for a real-world check.
The Gravity Knob
Gravity becomes a variable.
Microgravity research has an accessibility problem. You can’t just “do” µg on a Tuesday afternoon. Flight missions are rare, expensive and high-risk, and for most teams the gap between 1G on Earth and µg in orbit is a black box.
To put a number on “rare”: across all of 2024 the ISS National Lab supported 110 payloads across seven missions. Seven windows in a year, for every discipline competing for them. If your experiment misses one, you wait.
The Gravity Knob turns gravity into a tunable input. The twin models how a 3-D culture’s structure responds to the settling force of gravity: round, cohesive self-assembly when settling is removed; a flatter disk when settling dominates.
Be precise about what this is. Today the settling mechanism is validated against real 1G suspension-culture data, which shares the round, low-settling regime with true microgravity. That gives you a physics-grounded framework for reasoning about the 1G→µg transition. It is a bridge to space, not a substitute for flight data, and we don’t claim to have validated the microgravity-specific delta. Acquiring that validation data, against a matched 1G control, is the next step: we’re a partner on a ResearchSat-led ISS mission, funded through Round 3 of the SA Space Collaboration and Innovation Fund, that goes and gets it.
For the skeptics
We don’t ask for your trust. We show you the rigour.
In biotech, trust is earned through validation, not marketing. So here’s what’s under the hood:
- Reality-anchored. Calibrated against real-world biological datasets, not simulated data alone, to break circularity.
- Proven fidelity. In head-to-head tests, the surrogate reproduces its physics oracle at ~88% fidelity, and reaches 96% of the accuracy of a model trained directly on the real growth data, on a train-on-synthetic, test-on-real benchmark.
- Physics-enforced. Gated by hard physical invariants; it cannot suggest a biologically impossible result.
- Honest about limits. Every prediction ships with calibrated uncertainty, and we publish where the model cannot self-certify rather than hiding it. Microgravity is the clearest case.
- Edge-ready. The resulting model is lean enough to run locally on your own hardware, or on low-power hardware in orbit. Your data never has to leave the building.
Where teams put it to work
One validated cell line is the proven case today; the rest describe where the engine is designed to extend. We’ll always tell you which side of that line your problem sits on, and program-specific results are stated only with verified numbers.
Cancer & tumour biology
Explore 3-D biological growth across conditions you could never afford to run enough times in the lab.
Microgravity & space bio
Model µg behaviour from ordinary ground-based data, before (or instead of) a flight.
Bioprocess & cell culture
Narrow a huge parameter space down to the handful of runs worth doing physically.
Compound & condition screening
Rank candidates virtually, so the wet lab only ever sees the front-runners.
Frequently asked questions
What is a digital twin in biomedical research?
A working virtual model of a biological system. ALTDATA's twin pairs a high-fidelity physics simulator with a fast AI surrogate, so you can run experiments (growth, conditions, compounds) in software before, or instead of, the wet lab.
How is this different from a standard AI model?
Standard AI models learn patterns from past data and guess. Ours is anchored in physics and gated by physical invariants, so it cannot produce a physically impossible result. You get simulation rigour at AI speed, not a plausible-looking guess.
Can a virtual experiment replace wet-lab work?
No, and it isn't meant to. It replaces the wasted iterations. Run thousands of permutations virtually to find the handful worth testing, then confirm those in the lab. A confidence score on every result tells you when a real check is warranted.
What is the Gravity Knob?
It lets you treat gravity as a tunable input, modelling how a 3-D culture's structure responds to gravity's settling force. The settling mechanism is validated against real 1G suspension-culture data, which shares the round, low-settling regime with true microgravity, so it's a physics-grounded bridge for reasoning about the 1G to µg transition. It complements flight data; it doesn't replace it, and we haven't validated the microgravity-specific delta.
How do you know the twin is right?
On two axes, because one is not enough. We check the AI surrogate against the physics simulator that taught it, and we check it against real biological measurements, which is the test that stops a twin from only proving it agrees with itself. Hard physical invariants gate the output, and every result carries a confidence score telling you when to trust it and when to verify in the lab.
Is our data safe? Can it run on our own hardware?
Yes. The surrogate is small enough to run locally on your own hardware, so your data never has to leave the building.
Which research areas is it built for, and how do we start?
Today it's focused on biomedical research: tumour biology, microgravity and space bio, bioprocess and screening. The best start is a technical walkthrough on one of your own problems. Book a technical walkthrough →
The bottom line
Turn biological scarcity into a competitive advantage.
By turning limited datasets into scalable virtual experiments, ALTDATA helps biotech startups and research institutes discover faster, waste less R&D budget, and make scientific decisions with a level of confidence that was previously too expensive to reach.
Stop fighting the data gap. Start simulating it.
See it on one of your own problems.
The best start is a technical walkthrough. Bring a program you're working on and we'll show you the rigour of the engine.