The atlas — terpenes and flavorants

Every terpene and flavorant in the corpus, rotatable in 3D — with the clinic thread and the entourage evidence.

author: Denson Smith · part of The aroma molecules of cannabis

The atlas — the terpenes

Each card carries the live PubChem 3D structure — drag to rotate, scroll to zoom. Use the controls to change representation or spin them all. The colors are chemistry's standard code: gray spheres are carbon, white are hydrogen, red is oxygen — and where they appear, yellow is sulfur and blue is nitrogen.

Read the last line of every card: cannabis doesn't smell like cannabis so much as it smells like a blend of other plants — because it literally shares their aroma molecules. That shared chemical vocabulary is what a structured knowledge graph captures.

From a smell to the clinic — where β-caryophyllene points

Why hasn't the scrutiny happened? Not because the questions are uninteresting. Much of cannabis medical research is starved: attention and funding are scarce, and a long-running stigma has kept serious money and serious careers away from questions this size for decades. CB2 leads to one of the field's most-hyped clinical questions — does cannabis help dementia? — so we read the citation graph of that literature to see what it actually rests on. The honest version:

• The field stands on three pillars. A receptor-pharmacology bedrock (the CB1/CB2 cloning papers, ~5,000 citations each); a large body of preclinical neuroprotection work (cannabinoids protect neurons — in mice and cells, e.g. Hampson 1998, Ramírez 2005); and a thin but real clinical line on agitation.
• The only thing with real clinical support is symptom control — agitation. A clean lineage runs Volicer 1997 → van den Elsen 2015 → Herrmann 2019 (a nabilone RCT) → the Johns Hopkins dronabinol RCT (run 2017–2024; Rosenberg et al. 2026): well tolerated, and significant on one of its two co-primary agitation measures (PAS, −0.74/week vs placebo, p=.015) while the other (NPI-C) missed (p=.094). It matters because the standard alternative — antipsychotics — carries black-box stroke/death warnings in the elderly.
• The disease-modifying dream is still in mice. The "cannabinoids slow Alzheimer's" excitement rides on that same 1998–2007 preclinical work, with no human trial behind it yet.
• On risk: heavy/disordered use is a marker of higher dementia risk, but the rigorous genetic (Mendelian-randomization) evidence finds no causal link for moderate lifetime use — pattern and dose dominate.

The point for this document: a single aroma molecule — the smell of black pepper — opens onto an entire clinical literature, because it happens to also be a drug. Mapping those cross-domain reaches, with every claim grounded and the hype separated from the evidence, is the work. (The citation-graph structure above is verified metadata; the recent-trial figures are from grounded web search, not independently checked against the source papers.)

→ Read the full cannabis-and-dementia citation analysis (the three-pillar breakdown)

The trace modifiers — the beyond-terpene flavorants

Terpene profiles are remarkably similar across cultivars that smell nothing alike; the exotic notes come from these trace flavorant classes, each under 0.05% of the flower's mass. How they rewrite the whole percept — and whether they stop at the nose — is told on the story page.

The trace modifiers

Each card leads with the compound class, then its example molecule — drag any to rotate; the view/spin controls above drive these too.

The entourage effect — honestly

The popular claim is that terpenes and cannabinoids act synergistically — that the whole plant beats the isolated molecule. The evidence is more careful than the marketing:

• Broad synergy is mostly unproven. A Nov-2024 systematic review (André et al., Pharmaceuticals) found individual terpenes have real effects, but the claim that they synergistically enhance cannabinoid efficacy is clinically unproven for most combinations — commercial claims outpace the trial data.
• But one specific pairing is real. An April-2024 Johns Hopkins double-blind RCT found d-limonene significantly reduced THC-induced anxiety and paranoia — without dulling the high. First concrete human evidence that a terpene can selectively buffer a cannabinoid's side effects.

Where the molecule map ends and the expert begins

This atlas is the part a machine builds cleanly: molecule → aroma → botanical source → receptor, every identifier real and checkable. What it can't finish is the interaction — how the trace flavorants bend the terpene base, better or worse, and which of those blended profiles people actually prefer. That is a sensory-panel + preference-modeling problem (precisely what the Oswald 2023 study used a human panel for), plus validating compound effects in human cell models. The map is built; reading what it means for preference needs a domain expert.

→ The story: the aroma molecules of cannabis — same atoms, different smells, different effects