Rayca Science Molecular intelligence
One loop, inside your institution: evidence is resolved into a record, attrition becomes constraints, models and physics test each idea, and your scientists decide what is made.
Lead optimisation generates more evidence than a project team can hold. Too little of it shapes the next design.
Where R&D loses time and evidence today, and how Rayca addresses each point.
Fragmented evidence
Measurements sit across registries, plate exports, structural files, contract research reports and minutes. They are rarely reconciled into one decision-grade record, so each cycle is designed on a partial view of what the program has already measured.
How Rayca addresses it
One program record
Every measurement normalised, linked to its source and protocol, with conflicting results preserved rather than averaged.
Attrition that is not retained
The reasons compounds are discontinued are documented in reports but seldom translated into design rules. Liabilities characterised in one series can recur in the next.
How Rayca addresses it
Constraints with their evidence
Each supported liability becomes an explicit constraint, carrying its evidence, applied to every new design.
Models that do not transfer
Models trained on public or historical data often generalise poorly to a program’s own chemical series and rarely state their confidence, so a reliable prediction and an extrapolation look the same.
How Rayca addresses it
Program-trained models with uncertainty
Models learn from the program’s record, report how sure they are, and show where the evidence is thin.
Scarce synthesis and assay capacity
Design and make cycles are long and costly, yet candidates frequently reach the bench without rigorous physical evaluation of the hypothesis behind them.
How Rayca addresses it
Physics first, decisions with scientists
Hypotheses are tested in order of computational cost, and each cycle ends with a traceable recommendation your team approves, changes or rejects.
The program record 01
Decisions start from the full evidence of the program, with its provenance intact.
Resolving the record
Registries, assays, structures, reports and notes are read where they live.
Attrition into constraints 02
Liabilities a program has already paid to discover stay discovered.
From attrition to constraint
Program series
Trace the attrition
Each discontinued compound is linked to the measurements behind the decision.
Design constraint
Avoid this group at this position
New designs
Program-trained models 03
Predictions a project team can weigh, because their limits are visible.
Program-trained models
- Train on the record
- Predict with uncertainty
- Show thin evidence
- Refresh each cycle
Each endpoint model learns from the program’s own measurements.
Physics before synthesis 04
Synthesis effort goes to molecules that have already passed a physical test.
Physics in order of cost
Docking
Fast, broad
Molecular dynamics
Slower, follows motion
Free energy
Costly, decisive
Dock
Fast pose generation for the candidates that pass the constraints.
Agents and decisions 05
Every recommendation can be traced, questioned and overruled.
Agents, critic, decision
A planner breaks the program question into steps for specialist agents.
The loop closes 06
Each cycle begins from more than the last one did.
Closing the loop
This cycle
- Experiments run with your approval
- Results join the record
- Models retrained, constraints revised
- Design starts further ahead
Run the experiment
In your laboratory or through an approved partner laboratory.