About Helix7

GPCRs are the largest family of cell-surface receptors and the target of roughly a third of all approved drugs. Yet the data needed to reason about a given receptor — its structure, what binds it, where it is expressed, and what disease it drives — lives in half a dozen specialist databases with different identifiers and formats. Helix7 reconciles those sources on a common receptor spine so that a single page answers, for any receptor: what is it, what acts on it, where does it act, and why does it matter therapeutically?

Every human receptor in GPCRdb is included — 805 in total, spanning classes A (Rhodopsin), B1 (Secretin), B2 (Adhesion), C (Glutamate), F (Frizzled), the Taste 2 family, and the two odorant classes. Both the classic A–F nomenclature and the GRAFS grouping are stored, and colour is used consistently across the site to mean class.

Two tiers are treated differently. The 398 non-olfactory receptors receive full treatment. The 407 olfactory receptors are included for completeness and for the phylogeny, but carry sparse pharmacology and are hidden from browse by default.

93 receptors are flagged ORPHAN. These are the receptors that GPCRdb places in its dedicated orphan families — receptors for which no endogenous ligand is yet known. They are among the most interesting targets precisely because they are under-explored, and the browse page has an “Orphans only” filter to surface them.

Each receptor lists every experimental structure solved for it in the PDB, sorted by resolution, with method and activation state. The 3D viewer renders the best-resolution structure as an interactive cartoon; when no experimental structure exists, it falls back to the receptor’s AlphaFold model, so every receptor shows something. The chain is rainbow-coloured N→C so the seven transmembrane helices read clearly.

Ligand interactions are integrated from three sources: the Guide to Pharmacology (expert-curated, with clean mechanism-of-action typing), ChEMBL (comprehensive quantitative medicinal-chemistry data, target-mapped to the human receptor), and PubChem BioAssay (additional target-mapped assays and deposited screens). Where a source classifies it, each ligand is typed by pharmacological action — agonist, partial agonist, antagonist, inverse agonist, and allosteric modulator. Affinities (Ki, IC50, EC50, Kd) are normalised to nanomolar, and each measurement links to the PubMed paper it was reported in. Because these sources overlap, ligands are de-duplicated across them by InChIKey and name, and equivalent measurements are merged — so counts reflect distinct compounds, not repeated records. The potency strip plot shows every quantitative measurement on the −log₁₀(M) scale, where higher means more potent. The atlas currently holds 173,948 interaction records across 127,314 distinct ligands.

Tissue expression is the Human Protein Atlas consensus RNA dataset (normalised nTPM across 50 tissues). On each receptor page the values drive two linked views: a stylised anatomical figure whose organ systems are shaded by peak expression, and a ranked bar chart of the top tissues. Tissues are rolled up into organ systems (nervous, cardiovascular, digestive, and so on) so a receptor can be browsed by the system it acts in.

Disease links come from the Open Targets Platform, which integrates genetic, clinical, literature, and animal-model evidence into a single association score per target–disease pair. Helix7 shows each receptor’s top-scoring diseases with the evidence datatypes behind the score, plus the therapeutic areas they fall under. Across the atlas, receptors connect to 2,544 distinct diseases. Curated single-variant data (ClinVar/gnomAD) is a planned addition; today the genetic signal is captured in the association scores.

Every receptor carries a 0–100 development-interest score, designed to be transparent rather than a black box. It is a weighted composite of five signals: literature volume (30%, log-scaled so a few famous receptors don’t dominate), distinct ligands (20%), approved drugs (20%), solved structures (15%), and strength of disease evidence (15%). The receptor page shows the full breakdown.

Each receptor is also assigned a Target Development Level in the Pharos convention: Tclin (an approved drug acts on it), Tchem (a potent ligand exists), Tbio (biology known, no potent ligand), and Tdark (essentially unstudied).

Helix7 aggregates public data and preserves attribution on every record. Each receptor page carries source links, and the underlying databases are listed below — see the Citation & licenses page for the full licences and required citations.

A reproducible Python pipeline builds the atlas in stages: it seeds the receptor list and structures from GPCRdb, enriches with expression from the Human Protein Atlas, adds ligands and assays from the Guide to Pharmacology, layers disease associations from Open Targets, computes literature counts from PubMed, derives the development-interest score, and loads everything into MongoDB with a text index for search. Every network response is cached, so the build is reproducible and re-runs are cheap. The data is refreshed periodically as the sources update.

Helix7 is a research and target-triage tool, not clinical guidance and not investment advice. Affinity values are aggregated across assays of differing design; the potency plot is best read as a distribution, not a single canonical number. Absence of data — no listed ligand, no structure — reflects the current state of the source databases, not a biological certainty. Always consult the primary references, which are linked throughout.