Most of us take medications for years without ever seeing what's actually happening at the molecular level — the exact shape of the compound our body is responding to. Pharmacologists have that view. Almost no one else does.
That gap isn't just inconvenient. It's part of a larger problem: trust in medicine and science doesn't erode because people are given information — it erodes when they aren't. A prescription that arrives with no explanation anyone can actually follow, or a claim that can't be traced back to where it came from, makes skepticism the reasonable response, not a failure of the person asking questions.
Trust doesn't erode because people are given information. It erodes when they aren't.
LIGAND exists on the opposite premise. Every structure, mechanism, and indication on this site traces back to a public record — the same databases physicians, pharmacologists, and regulators actually use — so nothing here asks you to just take our word for it. Accuracy and accessibility were never really in tension; they just needed to be written as two honest versions of the same fact, instead of one compromise between them.
It's a work in progress — 199 compounds today, built to grow. But the standard doesn't move: if we can't point to where something came from, it doesn't go on the page.
LIGAND was built assuming its readers wouldn't be the same person twice. A pharmacologist checking a mechanism needs the clinical language and the primary source. Someone who just picked up a prescription needs the same fact stated in words that don't require a decade of training to parse — not a dumbed-down version, just a differently-aimed one. Every mechanism and indication on this site is written both ways, side by side, because accuracy and accessibility were never actually two different products.
That's useful past the individual reader, too. A clinician or pharmacist can point a patient to the plain-language explanation of their own prescription instead of reaching for an analogy on the spot — a second way to communicate a therapy accurately, not a replacement for the conversation. A student can treat it as a real reference, not a simplified stand-in for one. The underlying facts don't change depending on who's reading; only how much explaining they need does.
The seed set — 199 compounds across major therapeutic areas — was hand-picked and reviewed for pharmacological accuracy. Structure identifiers were never hand-entered: PubChem CIDs and SMILES, RCSB PDB entries, and UniProt accessions are all fetched programmatically from PubChem's PUG REST API and RCSB PDB's Search and Data APIs. Where a biologic's own engineered sequence has no public structural record, UniProt fills in its molecular target instead of leaving the field empty.
Small molecules are rendered with 3Dmol.js from real 3D SDF conformers pulled live from PubChem. Biologics are rendered with NGL from real deposited structures pulled live from RCSB PDB. Nothing on the visualization panel is an illustration or a stock image — if a structure renders, it's the actual compound.
Next.js 16 (App Router) and Tailwind CSS v4, in TypeScript throughout. No backend — the dataset is static, enriched offline by a Python script, and shipped with the app.
The color system is functional, not decorative — teal for small molecules, violet for biologics — paired with shape and text labels so it never depends on color alone. The clinical / plain-language toggle exists because accuracy and accessibility aren't actually in tension; they just need separate copy.