ELYRIS / BIOPHYSICAL INTELLIGENCE
BIOPHYSICAL INTELLIGENCE FOR DRUG DEVELOPMENT.
Computational twins that show what an asset will do, why it will do it, and whether it deserves the next experiment.
LURA is our oncology biophysical twin engine. Test an asset across genetic context, spatial delivery, and mechanistic response — then export decision-ready, governed evidence.
01 / THE PROBLEM
Drug programs learn
the truth too late.
Biological evidence is fragmented. Spatial behavior is hard to observe. Mechanism gets reconstructed across disconnected assays. Failures are often understood only after another experimental cycle — or much later in development.
Fragmented evidence
Teams reconcile partial assays, disconnected datasets, and competing narratives before they can decide what to run next.
Spatial blindness
Delivery, stroma, vasculature, and hypoxia decide whether an asset reaches the cells that matter — and flat models miss them.
Late failure
The cost is not “simulation is slow.” It is capital and years spent learning truths that could have been interrogated earlier.
What if you could interrogate this before the next experiment?
02 / THE LOOP
SCREEN BROADLY. INTERROGATE THE BIOLOGY. EXPORT THE DECISION.
LURA shortens the path from hypothesis to next experiment. Not a leaderboard — a picture of what happens and why, packaged for review.
SCREEN / 01
Which assets deserve the next cycle?
Rank a library against the biology you actually care about. Promote only what survives genetic and contextual selectivity.
03 / BIOLOGY
Three questions
every asset must answer.
PGx
Pharmacogenomics
Does this asset work in the biology you are actually targeting?
Genetic context. Genotype-aware ranking, response heterogeneity, and the genetic determinants that separate responders from non-responders.
PK
Pharmacokinetics
Can the drug reach the cells and compartments that matter?
Spatial exposure. Penetration, vasculature, stromal barriers, hypoxia, and spatial concentration inside tissue — not a dish average.
PD
Pharmacodynamics
What happens to the tumor when the drug gets there?
Mechanistic response. Target engagement, pathway behavior, resistance, and downstream effects under kill pressure.
One asset. Three biological questions. One continuous biophysical twin — so genetics, delivery, and response are read together, not as three disconnected reports.
04 / DECISIONS
What changes when
Elyris exists.
LURA is decision infrastructure for pharma — not interesting simulation software. These are the calls teams actually need to make.
Go / no-go
Is there enough mechanistic and spatial support to advance this asset?
Repurposing
Where else could this mechanism work?
Combination strategy
What intervention could overcome the observed resistance?
Experiment design
What is the highest-information experiment to run next?
Portfolio allocation
Which programs deserve another dollar and another quarter?
05 / EVIDENCE
RECEIPTS,
NOT ADJECTIVES.
A result should not just be impressive. It should survive review. The Decision Evidence Pack carries every conclusion with its inputs, model context, provenance, assumptions, uncertainty, and supporting outputs.
What happened
Pharmacogenomics fit, pharmacokinetic delivery, and pharmacodynamic response — structured so science and leadership read the same picture.
Why we believe it
Provenance, model context, and assumptions bound to the artifact — not a slide that loses its chain of custody.
Where it can break
Uncertainty and degradation indicators travel with the claim. Fail closed when evidence does not clear the bar.
DECISION
EVIDENCE PACK
EVIDENCE PACK
- ASSET IDENTITY
- RESOLVED
- PHARMACOGENOMICS
- BOUND
- PHARMACOKINETICS
- STRESSED
- PHARMACODYNAMICS
- TRACED
- UNCERTAINTY
- FLAGGED
LGOVERNED
EVIDENCE
EVIDENCE
06 / SCALE
TEST THE
POSSIBILITY SPACE.
Hypothesis → biophysical twin → thousands of interventions → mechanistic picture → governed evidence → better next decision. Built for campaign-scale interrogation, not one experiment at a time.
drug–cell pairs
1,699 cells × 1,545 compounds
Why it matters: enough of the space to shortlist with confidence, not a handful of lucky hits.
campaign screen
A100 live receipt
Why it matters: library-scale ranking at a speed that changes when you ask the next question.
cross-dataset AUROC
PRISM train → GDSC2 holdout
Why it matters: signal that generalizes beyond the training distribution.
scaffold AUROC
held-out Oracle panel
Why it matters: useful on chemotypes the model has not memorized.
COMPUTE FABRIC
NVIDIA A100 / H100 CLASS
PARALLEL SCREENING
GOVERNED EXPORT
07 / ELYRIS
Biology should become more testable
before it becomes more expensive.
Elyris builds biophysical intelligence for drug development. LURA begins in oncology. The larger mission is to make biological decision-making faster, more mechanistic, more auditable, and more capital-efficient.
ElyrisBiophysical intelligence company
LURAOncology biophysical twin engine
SimulationComputational substrate
Decision Evidence PackEnterprise artifact
OutcomeBetter, faster drug decisions
BUILT FOR
THE WORK.
Production paths for modern GPU fleets and enterprise cloud. Program membership is infrastructure access — not scientific validation of LURA.
08 / ACT
BRING US
AN ASSET.
A stalled program. A difficult go/no-go. A repurposing hypothesis. A mechanism you need to understand.
↗