Being wrong should cost a second, not a decade.
Elyris Labs is a computational oncology company in Fort Worth, Texas. We build LURA: an engine that screens a compound against genetic context, moves it through tissue in three dimensions, and returns the reason it wins or fails. All of it is aimed at one number — what it costs a programme to discover it was never going to work.
The vision
We think the way a drug is chosen is about to change completely, and that the change is computational rather than biological.
Today a candidate is picked on a plate, defended in an animal, and tested in a person, and each step is slower and more expensive than the one before it. The information that would have stopped a bad programme almost always existed earlier than the step that finally revealed it. It was simply never computed, because until recently it could not be.
Elyris Labs exists to compute it. We build the layer where a compound, a genotype and a tissue can be put together and interrogated before a programme commits to any of them — and where the answer that comes back is one a scientist is able to argue with.
The problem
Around nine in ten candidates that enter clinical development never reach approval. Carrying every one of those failures, the cost of bringing a single drug to market is counted in billions of dollars and well over a decade.
Almost none of that is the price of discovering that something works. It is the price of discovering that something does not — years after the decision to pursue it, and long after the evidence that would have predicted it was available in principle.
The paradigm shift
In December 2022 the FDA Modernization Act 2.0 removed the statutory requirement that a drug be tested in animals before it can enter human trials, and named computational modelling among the alternatives that may stand in place of it.
That was a change in what is allowed to count as evidence. It did not, by itself, produce any. A computational result earns its place only when it is built to a standard a reviewer can audit: a stated method, a stated uncertainty, a provenance chain, and an unbroken line between what was measured and what was inferred.
We are built for that standard rather than merely for the opening it created. Every result LURA returns carries the engine that made the claim, the assumptions under it, and its confidence on the face of it. A computational answer nobody can challenge is not evidence — it is an assertion with arithmetic attached.
The product
LURA runs parallel virtual tumours on GPU.
It screens a compound against genetic context at a scale that used to mean a screening campaign — 2.5 million drug–cell pairs evaluated in 1.5 seconds on a single NVIDIA A100 — and then does the thing a screen cannot do at all. It solves transport, gradients and pressure in three dimensions, so the tissue a compound never reaches becomes part of the answer instead of being absent from it.
What comes back is not a score. It is a ranked, causally explained shortlist with its provenance attached, in seconds rather than months.
- Systems exist that predict response at scale.
- Systems exist that simulate tissue physics.
- We do not know of another that does both and returns them as a single governed, reviewable answer — which is the only form a decision can actually use.
The compounding advantage
Catching a failure in 1.5 seconds rather than six months does not save six months. It changes what a portfolio is.
A programme that dies early hands its budget, its bench time and its people back to the pipeline while all three are still worth something. Do that once and it is a saving. Do it across every programme, every quarter, and capital efficiency stops being a line item and starts to compound — in the way interest compounds, and for the same reason.
That is the whole argument. Not that we are better at prediction in the abstract, but that being wrong quickly and cheaply, over and over, is worth more than being right slowly.
The team
We are a specialised task force of computational, biological and complex-systems engineers. We do not write software about biology. We design digital twins, solve biophysical transport, and run the secure infrastructure a pharmaceutical company is willing to put its chemistry on — which is why the people who built it come from all three disciplines rather than one.
End-to-end architect of LURA's 3D physics loops, Graph Laplacian solvers and parallel GPU orchestration. A systems engineer who builds digital twins from scratch.
Former Global Health Fellow at Rice360. Microphysiological systems and transport modelling. The scientific anchor between what the code claims and what a wet lab can show.
Complex technical project management and industrial operations, previously at American Electric Power. Turns mathematical modelling into enterprise pipelines that hold under load.
Cloud architecture and distributed multi-agent control systems. Architect of the zero-trust, multi-tenant deployment layer built for Tier-1 pharmaceutical data isolation.
Twenty years of clinical engineering and global regulatory practice, holding LURA to the international validation standards it will eventually be read against.
The company
- Legal entity
- Elyris Labs Inc.
- Headquarters
- Fort Worth, Texas, United States
- Enquiries
- info@elyrislab.com
- Confidentiality
- How we handle your data