Controlled autonomous scientific systems
Autonomous
peptide
discovery
infrastructure
Protean Labs builds proprietary runtime systems for constrained peptide optimization, failure-aware ranking, bounded adaptation, and scientific execution at infrastructure scale.
Runtime orchestration
Scientific execution under control.
01
source lineage
02
failure motifs
03
ranked candidates
04
review artifacts
Architecture
Autonomous scientific infrastructure for reviewable discovery.
Protean Labs is the operator of a proprietary discovery engine. The public surface communicates systems architecture, not internal operating mechanics: evidence modeling, constraint synthesis, validation, ranking, learning, and review.
Proprietary discovery engine
Protean Labs operates a purpose-built engine for constrained peptide optimization and scientific orchestration.
Controlled autonomy
Autonomous cycles are bounded by deterministic gates, scoring contracts, and reviewable system state.
Bounded learning systems
Adaptive behavior improves prioritization without uncontrolled training loops or open-ended feedback amplification.
Failure-aware optimization
Negative evidence, failed motifs, and rejection patterns inform the engine instead of disappearing from the process.
Execution cycle
Evidence becomes controlled motion.
Every cycle moves through explicit control points: evidence intake, constraints, generation, validation, scoring, explanation, bounded adaptation, and research packaging.
Evidence intake
Scientific records and negative signals enter a curated evidence plane.
source lineage
Constraint synthesis
Sequence rules are composed into a bounded optimization surface.
design bounds
Candidate generation
Autonomous proposal systems explore controlled peptide design space.
candidate field
Validation gates
Deterministic checks reject invalid, unstable, or poorly formed candidates.
gate state
Ranking
Multi-factor scoring ranks candidates against constraints and failure memory.
priority vector
Interpretability
Rationales expose why a candidate advanced, stalled, or was rejected.
review trace
Bounded adaptation
Learning loops adjust within caps without rewriting the scoring contract.
controlled update
Research package
Reviewed outputs become structured packages for scientific and IP evaluation.
handoff layer
Autonomous runtime walkthrough
One bounded cycle from evidence to review.
Protean operates as a controlled scientific runtime: evidence enters, memory forms, hypotheses shape computational experiments, candidates move through deterministic gates, and each cycle becomes reviewable research state.
50
Ranked candidates
latest bounded cycle
15
Candidate explanations
top review set
6
Hypotheses
reviewable research questions
43
Sequence clusters
mapped candidate field
active stage 01
Scientific Data Acquisition
Literature, evidence records, failures, patents, peptide databases, and planned assay or structure sources enter as provenance-bearing records.
A scientific runtime is only as credible as its source discipline. Protean treats source origin, freshness, duplication, and contradiction as part of the research state.
input
papers, protein records, failure signals, patent context
control
source trust scoring, deduplication, bounded ingestion
output
curated evidence plane
runtime causality map
boundedcycle contract
Operating model
Autonomy with explicit control surfaces.
Models support proposal and interpretation. Deterministic validators, scoring contracts, failure memory, and bounded learning remain the control plane.
Runtime mode
continuous
Gating
deterministic
Learning
bounded
Failure memory
first-class
Model layer
orchestrated
Evidence plane
curated
evidence → constraints → candidates → gates → ranking → review
Candidate field
Ranking as an execution surface.
PX-41
lane 01
advance
constraint fit
0.82
PX-33
lane 02
hold
failure proximity
0.67
PX-28
lane 03
review
stability signal
0.74
PX-19
lane 04
reject
gate conflict
0.31
Runtime proof
Reviewed public export, not a dashboard demo.
The candidate explorer is a public proof element for a proprietary research runtime: evidence enters, constraints shape the candidate field, deterministic gates reject weak sequences, and top candidates move toward review and experimental handoff.
Failure motifs continue influencing rank instead of disappearing after rejection.
Bounded adaptive learning adjusts prioritization conservatively.
Selected candidates generate review artifacts before wet-lab decisions.
candidate lane
ranked
01
ingest
Source records enter with provenance, freshness, and trust context.
evidence intake
02
extract
Literature, failure, and peptide signals normalize into structured evidence.
typed signals
03
constrain
Cleavage risk, novelty pressure, synthesis bounds, and failure proximity shape search.
design bounds
04
generate
Candidate families are proposed inside the bounded optimization surface.
proposal set
05
validate
Deterministic gates screen residue validity, motif burden, and warning load.
gate state
06
rank
Multi-axis scoring prioritizes stability, practicality, novelty, and risk.
priority vector
07
learn
Bounded updates preserve failure memory while adjusting prioritization conservatively.
controlled delta
08
handoff
Top candidates become review packages for assay planning and scientific decision-making.
review packet
Provenance
Private science with public proof.
Protean’s provenance layer prepares cryptographic commitments, lineage summaries, and Base-prepared proof objects without exposing proprietary sequences or unfiled research detail.
Explore provenancePrivate vault
raw science
Artifact layer
redacted proofs
Base registry
hash anchors
Public proof records demonstrate artifact integrity and lineage. They do not disclose candidate payloads or imply biological validation.
Infrastructure
Systems for constrained optimization, failure memory, and reviewable output.
Evidence plane
Curated scientific records, source traces, and negative signals feed the optimization system.
Constraint engine
Design rules shape candidate space before scoring, ranking, and review occur.
Model orchestration
Specialized model routes support proposal, interpretation, and feature extraction.
Scoring architecture
Multi-axis ranking balances stability, synthesis practicality, novelty, and failure proximity.
Failure memory
Rejection signals and degradation patterns become reusable optimization intelligence.
Research packaging
Candidate evidence is structured for internal review, experimental planning, and IP strategy.
Systems doctrine
Autonomous computation, bounded by scientific controls.
Autonomous proposal
Candidate systems explore peptide design space under explicit constraints.
Feature extraction
Sequence-level signals are converted into ranking and review context.
Runtime control
Orchestration keeps execution bounded, observable, and reproducible.
Model routing
Specialized routes support generation, interpretation, scoring, and evidence extraction.
Research systems
A proprietary engine with controlled public artifacts.
Research papers, candidate assessments, and documentation are downstream communication layers. The discovery engine remains Protean’s proprietary infrastructure.
Read research outputs
Scientific integrity
Claims remain downstream of evidence.
Protean treats evidence, constraints, rejection logic, and review thresholds as infrastructure. The system can prioritize candidates; it does not convert computational confidence into biological truth.
Computational rankings are research prioritization signals.
Experimental claims require controlled assay evidence.
The platform is built to preserve constraints, provenance, and rejection logic.
Documentation
Protocol-grade notes for the scientific operating layer.
The docs explain Protean’s platform concepts, ingestion method, model layer, optimization principles, bounded learning posture, and research pipeline with public-safe technical depth.
Enter docsRoadmap
Infrastructure expands only where the science can hold it.
01
Current
Autonomous peptide infrastructure
Continuous orchestration, constraint-guided generation, failure-aware ranking, and bounded adaptation.
02
Expanding
Deeper research memory
Broader negative-evidence modeling, richer assay feedback pathways, and stronger candidate lineage.
03
Protocol
Infrastructure alignment research
Protean is researching reviewed collections, public-safe provenance, and controlled contribution workflows as long-term infrastructure alignment surfaces.
04
Research
Experimental translation
Computational prioritization remains connected to controlled experimental follow-up and scientific review.
Protocol research
Future infrastructure alignment
$PRTN is Protean’s protocol-scale infrastructure layer for reviewed collections, verifiable provenance, and public-safe coordination surfaces.
Future participation pathways, if pursued, would be separate, review-gated, and subject to applicable requirements.
Explore $PRTN architecture