semantic pre-filter for discovery

Patient biology before wet lab.

Encode disease mechanism, mutation, pathway context, and validation constraints into a deterministic field. ARBITER ranks candidate compounds by mechanistic coherence — not affinity, not marketing.

open screening console view ranked audit
103
compounds screened
22
above threshold
72D
meaning geometry
26MB
deterministic engine
coherence screening · patient-derived field
LRRK2 G2019S / lysosomal-autophagy dysfunction / dopaminergic rescue
A semantic-mechanistic filter that screens compounds against patient state, pathway dysfunction, and validation constraints before organoid or wet-lab work.
Patient profile encoded
Mechanism constraint field active
00patient state
01mechanism field
02candidate set
03coherence rank
04validation
translational chain

From patient state to ranked intervention

ARBITER does not predict binding or clinical outcome. It calculates semantic-mechanistic fit: how coherent each candidate is with the patient-derived disease field and biological constraints.

00
Patient profile
Mutation, pathway dysfunction, cell phenotype, clinical objective, validation model.
01
Constraint field
Target biology, mechanistic requirements, disqualifiers, validation rigor.
02
Candidate library
Compounds, repurposing evidence, mechanism notes, and known liabilities.
03
Coherence scoring
ARBITER ranks fit against the patient-mechanism field.
04
Above threshold
Only the strongest candidates advance to organoid or wet-lab validation.
01 · patient & mechanism
disease-derived constraints
02 · candidate library
upload CSV or use sample
drop CSV / text file
one compound per row · first column = name
ARBITER · coherence console
ready
Load the LRRK2 sample to assemble the patient-mechanism query.
screening progress
0/0
threshold ⩾ 0.70 → advance to validation
0
candidates passing filter
ranked candidate audit
deterministic coherence
No screen executed score
#candidatescore

measured

Semantic-mechanistic coherence between patient-derived field and compound description.

not measured

Binding affinity, PK/PD, toxicity, trial success probability, or medical recommendation.

deterministic

Same input → same output. Calculator, not a learner. Fully auditable.

research-grade prototype · not a clinical decision system · all outputs require wet-lab validation, pharmacological review, and safety review.