Generate, filter, test.
Sequence models prioritize fold, expression, brightness and stability. The spin response is usually discovered later by trial and error at the bench.
Rank candidates for quantum biosensor testing.
Nebula helps teams building spin-linked protein biosensors decide what deserves bench time first.
The bottleneck
The problem is not generating more sequences. It is knowing what deserves the bench.
The missing pre-bench question: which candidates have intrinsic spin-linked sensing potential worth testing?
Prioritize before the bench.
Keep the developability filter. Add a spin-physics layer. Rank candidates that are practical enough to build and promising enough to test.
Nebula engine
Two gates before the bench.
A candidate rises only when it is thermostable enough to advance and spin-active enough to justify testing.
Gate 01
Thermal stability.
AstraDTM and developability signals filter for candidates that can plausibly survive real development.
Estimate thermal stability context before a novel scaffold consumes assay and construct bandwidth.
Use Orbion model signals for fold, expression risk, scaffold constraints and cofactor context.
Only candidates with enough practical protein context move forward to spin-physics scoring.
Gate 02
Spin activity.
Nebula scores the intrinsic spin physics that sequence-only models cannot infer.
Resolve whether the candidate contains a plausible spin-active center: radical pair, triplet state, metal site or cofactor.
Model whether the spin event can couple to an optical readout strongly enough to be worth testing.
Advance only candidates that pass both gates: thermostable enough to develop and spin-active enough to test.
The proven foundation
Before spin, a biosensor still has to be developable. Orbion's models handle the classical protein layer: fold, stability, expression risk, binding context and manufacturability-relevant liabilities. Nebula adds the spin-linked ranking layer on top.
01
AstraROLE2 and AstraSUIT2
Annotate protein function and estimate target suitability.
Read the paper02
AstraBIND
Predict likely ligand-binding residues so design workflows can flag sites that should be treated cautiously.
Read the paper04
Developability signals
Predicted thermostability, folding stability, disorder and expression-risk signals, scored before bench testing.
The proven foundation under Nebula: selected Astra methods are documented in bioRxiv preprints with reported benchmarks. The spin-linked Nebula layer is the new layer now opening for design-partner evaluation.
The partnership
Design partners shape Nebula's first release.
01
Run on the proven foundation
Start from developability signals predicted by our models before Nebula computes the spin layer.
02
Shape the workflow
Help define the evidence views, ranking logic and shortlist formats real biosensor teams need, with room for collaborative case studies and launch marketing around Nebula as a cutting-edge discovery engine.
03
Your IP stays yours
Your targets, program IP and evaluation outputs stay yours under written partner terms.
04
Early access, heavy discounts
Design partners get priority access to partner runs and materially discounted early pricing for helping shape the product.
The first design-partner window is limited. Book a call now.