!
Biology and interface assumptions are separated.Documented odor-response candidates feed a small project-defined circuit for visualization. Slider values are relative inputs—not ppm, sensory thresholds, or bottle measurements.
Interactive laboratory
Compose. Propagate. Intervene.
Begin with a teaching profile or move each compound manually. Save two circuit states to compare them, silence one input pathway, or replace the downstream weights with a deterministic rewired control.
Reduced circuit view
ORN input → glomeruli → projection neurons → Kenyon-cell groups → engineered readouts
Leading channel
Sample A
No sample saved yet.
Sample B · distance —
No sample saved yet.
Read before interpreting
What the model is—and is not
The educational value comes from exposing the translation layer. Every number after the documented odor response is a project decision that can be inspected or challenged.
Documented layer
Candidate odorant responses and named glomerular pathways, with strongest confidence for geosmin through Or56a and DA2.
Constructed layer
A small deterministic PN-to-Kenyon matrix, normalization rule, ethanol gain modifier, and three display readouts. These are not measured fly behavior.
Control layer
An input-free baseline, single-pathway silencing, and seeded rewiring make it possible to ask whether a visible result depends on the intended structure.
No hidden wine score. “Fruit,” “floral,” “alert,” and “signature energy” summarize this interface’s internal state. They do not measure deliciousness, typicity, faults at a sensory threshold, health effects, or human preference.
Mapping register
Where each input enters
Confidence labels apply only to the first biological mapping—not to the downstream teaching circuit.
| Input | Wine context | Fly entry used here | Confidence | Interface treatment |
|---|---|---|---|---|
| Geosmin | Earthy odor / environmental taint context | Or56a → DA2 | High | Dedicated channel |
| Ethyl acetate | Fruity to solvent-like, concentration dependent | DM1-led broad ester response | Medium | Distributed across DM1/DM2 |
| Isoamyl acetate | Banana / pear-drop ester | DM2-led ester response | Medium | Distributed across DM2/DM1 |
| Acetic acid | Volatile acidity | Ir64a-associated acid pathway / DC4 simplification | Medium | Acid channel; no threshold claim |
| Linalool | Floral terpene | Distributed terpene-sensitive response | Provisional | Pooled D / DL5 teaching channel |
| Ethanol | Fermentation matrix component | No single receptor assigned | Deliberate limit | Global gain modifier |
Open development
Inspect the work behind the work
The research ledger lists every GitHub project under review, every source queue, the non-duplication boundaries, and the claims that still need stronger evidence.
Research ledger
Review code provenance, scientific sources, and mapping confidence.
Current milestone
Interactive aroma composition, deterministic propagation, A/B comparison, blind profiles, silencing, and rewired controls.
Next evidence work
Replace provisional pooled pathways with concentration-aware response curves where direct primary evidence supports them.