GPCompaRe database

Paired active and inactive AlphaFold-Multistate models for 283 human Class A GPCRs, compared using residue-residue contact scores (RRCS). Each report links the structural result to GPCRdb generic numbering, conservation, AlphaMissense predictions, cross-family Core Functional Residue positions, and gnomAD variation for the 279 receptors that have it.

Browse 283 GPCRs Database statistics

Dataset and analysis

Receptors	283	human Class A GPCRs with a complete active and inactive pair
Receptor families	60
Structural models	566	two AlphaFold-Multistate models per receptor
Contact-pair records	213,456	residue pairs detected in either state
Threshold-passing changes	23,025	contacts exceeding their own receptor's threshold
Excluded	4	receptors without a complete active and inactive pair

Positions and contact pairs highlighted by GPCompaRe should be interpreted as model-derived structural candidates, not direct experimental evidence of function.

How to read a GPCompaRe contact

One real contact from the PAR2 report, from the topology diagram to the number.

Fragment of the PAR2 snake plot, residues shaded by their contact-score change,
red where contacts are stronger in the active model and blue where they are
stronger in the inactive model. M159 in TM3 and F300 in TM6 are ringed, labelled
with their generic numbers, and joined by a line marking the contact between them. — PAR2 (F2RL1), transmembrane helices 3 and 6, coloured by the report's own ΔRRCS view.

RRCS assigns a contact score to a residue pair from the interatomic distances within a structure. The same pair is scored in the active model and in the inactive model, and the difference is ΔRRCS: active minus inactive.

M159 and F300 as they appear in the PAR2 report
Residue 1	M159 `3.36x36`, TM3
Residue 2	F300 `6.48x48`, TM6
Active RRCS	0.67
Inactive RRCS	5.66
ΔRRCS	−4.99
PAR2 threshold	2.60, so this contact passes
Rank in PAR2	27th of 776 contact pairs by \|ΔRRCS\|, so not the largest change in this receptor

The contact is stronger in the inactive model, so it is blue. Contacts stronger in the active model are red.

F300 occupies the aromatic 6.48 toggle-switch position, one of the canonical activation landmarks. Both residues sit in transmembrane helices.

Open the PAR2 snake plot

Core Functional Residues across Class A GPCRs

Consensus Class A topology with each generic position shaded by the number of receptors showing a threshold-passing contact change, and the DRY, NPxxY, CWxP, PIF and D2.50x50 landmarks marked — Each position is shaded by how many of the 283 receptors show a threshold-passing change there. Lettered and outlined positions mark canonical activation landmarks.

A receptor-level Core Functional Residue (CFR) is a residue that participates in at least one contact whose absolute ΔRRCS passes that receptor's own threshold.

Mapping those receptor-level CFRs onto GPCRdb generic positions gives recurrent CFR positions across the family. A position is counted as recurrent when it is identified as a CFR in at least three receptors.

That mapping needs a GPCRdb generic number. The cross-receptor recurrence statistics reported here are the results submitted with the manuscript, computed on the full GPCRdb residue maps for all 283 receptors, so they are not limited by how much of a receptor its own annotation file numbers.

No motif was selected in advance. The cross-receptor analysis recovers established parts of the DRY, NPxxY and CWxP activation machinery on its own.

See the ranked Top 50 recurrent positions

What each receptor report contains

An interactive snake plot with selectable colour views and contact arcs
Contact pairs ranked by |ΔRRCS|, with the active RRCS, the inactive RRCS and the difference, up to 1,000 rows
The receptor's own threshold, and which contacts pass it
Cross-family Core Functional Residue positions mapped onto this receptor
GPCRdb generic numbering
Sequence conservation
gnomAD missense variation at contact-forming residues, for the 279 receptors that have variant data
AlphaMissense pathogenicity predictions
Ranked tables with CSV export

Use the search above to open a receptor. Example reports: ADRB2, PAR2.

Analysing your own structures

Alongside the database, GPCompaRe covers analysis of your own paired active and inactive structures, applying the same RRCS comparison used to build this database and writing standalone HTML and CSV output.

A releasable version of that program is not yet available. Analyses you run locally would stay local: they are not added to the public database.

Downloads

Methods and data sources

Structures are paired active-state and inactive-state models from AlphaFold-Multistate. Contacts are scored with RRCS. Generic numbering and segment boundaries come from GPCRdb. Population variation comes from gnomAD v4, pathogenicity predictions from AlphaMissense, and conservation from ProtVar and UniProt.

Each receptor has its own ΔRRCS threshold, computed as max(mean(|Δ|) + σ, 0.2).

566 models were analysed, two per receptor. Four further receptors listed in the Class A reference set are excluded because they lack a complete active and inactive pair: GPR26, LGR5, OPSX and RXFP1.

Version, downloads and citation

Release: GPCompaRe database release 2026.08
DOI: DOI pending release
Downloads: Database release and analysis software
Contact: Contact the lab

GPCompaRe was developed as the companion resource for the associated study. Corrections and substantial updates will be documented as new releases. No fixed update schedule is currently planned.