Discovery, not just screening
Topology-first modeling surfaces tractable targets and mechanisms that conventional pocket-based screening misses — including disordered and boundary-targetable proteins.
BioTwin · An 8Braid Inc. platform
Run the entire drug discovery program in silico — before the first wet experiment.
BraidEra simulates a drug program end-to-end across every resolution layer — from quantum and atomic structure up through pathway, cell, tissue and organism — and carries cryptographically auditable evidence at every step. Walk in with a target. Walk out with a virtual shelf-ready dossier that converts cleanly into an IND-enabling program. The lab becomes confirmation, not discovery.
Built on the 8DB evidence-native information substrate.
One persistent therapeutic program
Not a target list. Not a pocket score. A complete path to validation.
BioTwin composes the capabilities required to keep difficult programs moving—and preserves every result so the next decision starts from accumulated evidence.
Disease mechanismHidden tractabilityIntervention architectureCandidate designEngagementFunctional pharmacologyPK / efficacySafetyValidation + dossierProgram memory
Targeted protein degradation
The field has roughly 600 E3 ligases. It builds almost everything with two.
BraidEra maps noncanonical E3 recruiter opportunities before a team commits to protein production, screening and medicinal chemistry. A computational family-wide campaign screened 371 catalytic ligases plus 14 CRL adaptors, prioritized 243 candidates and generated 677 residue-level recruiter-anchor hypotheses.
These are computational starting points—not experimentally validated ligands or degraders.
371catalytic E3 ligases screened
243computationally prioritized
677anchor hypotheses generated
1evidence-native program substrate
Evidence you can audit
Every prediction carries an evidence tier, a confidence vector, and a falsifiable criterion. Nothing is asserted that can't be traced and challenged.
Cross-scale by construction
One coherent model spans quantum to organism, so a finding at the atomic layer propagates its consequences — and its uncertainty — all the way up.
How it works
Three moves, repeated until the dossier is shelf-ready.
-
1
Anchor to known science
Every layer is calibrated against published results and reference databases. Predictions start tethered to ground truth, with confidence that reflects the supporting evidence.
-
2
Simulate end-to-end
Propagate a candidate through every resolution layer. Where the model can do what conventional methods can't, it produces a specific, testable prediction instead of a vibe.
-
3
Expose and challenge the holes
Open questions are first-class. Each gap is tracked as accepted, assumed, or challenged — and a challenged gap auto-generates the experiment that would falsify it.
Reality is multimodal. No database is — so we built one.
Every other system makes you pick one shape and force the rest of reality to fit it.
A molecule, a regulatory pathway, a cell, and a whole organism are fundamentally different kinds of structure — and the science lives in the dependencies between them. Yet every data system today makes you choose a single model: rows, or documents, or a graph. Fragment reality across incompatible systems and those dependencies vanish in the seams, taking the discoveries hiding inside them along too. BraidEra runs on a computational substrate, built from scratch in Rust, that represents every known data modality natively in one engine — plus new ones, like topological braids. We don't flatten nature to fit the database. We simulate it as it actually is.
Every modality, one substrate
Quantum, chemical, structural, network, and organism-scale state all live as atoms in a single engine — not a graph database bolted onto a document store, with lossy hand-offs between systems that each see only part of the picture.
Modalities no one else has
Path-dependent biology — epigenetic history, regulatory sequence — is modeled as topological braids
that preserve the exact order of events. BraidQL operations like PULL_STRAND and
EXCISE_CROSSING ask questions that are physically impossible to write in SQL or GraphQL.
Confidence built into the atom
Every atom carries a 7-axis confidence vector at the storage layer itself. The engine refuses to answer beyond its confidence frontier — it surfaces what it doesn't know instead of fabricating it.
PULL_STRAND FROM strand::0 TOP 5
WITHIN braid::MYCN
RETURN strand_id;MYCN braid — in order, with the path preserved. There is no way to write this in SQL.Virtual shelf-ready means falsifiable
"Virtual shelf-ready" is only credible if every claim has a known-science anchor and a path to being proven wrong. BraidEra ties each claim dynamically to its supporting evidence; contradictions propagate through downstream confidence, and the system gets more confident — and more honest about its limits — as more validation lands.
- Provenance and reversibility on every result
- Per-claim confidence vectors, not single scores
- Auto-designed falsification experiments for contested predictions
- A living calibration record that compiles as you build
One model, every scale
From electrons to the whole organism — a single coherent substrate.
R0Quantum / electronic
R1Atomic / structural
R2Residue / segment
R3Domain / disorder
R4Complex / oligomer
R5Pathway / network
R6Cellular / phenotype
R7Tissue / ADMET
R8Organism / PK-PD
Bring us a target.
BraidEra is in active development with design partners. If you're working on a hard target — disordered, "undruggable," or resistance-prone — we'd like to hear about it.
Request early accessContact hello@8braid.com.