Astra Suite

AstraBIND

Where Ligands Bind — and What Binds

Function often comes down to a pocket. AstraBIND locates the residues that make up ligand-binding sites and retrieves the ligand types most likely to sit there — combining sequence, structure, and homology in a single pass.

Binding ResiduesLigand TypesSequence + Structure
Read the Preprint
Retrieved Ligands
Retinal · Small Molecule0.94
Zn²⁺ · Metal0.16
GDP · Nucleotide0.09
Illustrative Ligand Retrieval · a Receptor Pocket
Signature Strength

A 2-Million-Ligand Panel, from Sequence

AstraBIND groups binding residues into coherent pockets and screens them against a curated panel of over 2 million ligands in 17 categories — nucleotides, cofactors, metals, peptides and more — fusing sequence, structure, and homology in a single pass.

2M+
Ligands · 17 Categories

Real Structure · Abl Kinase + Imatinib (PDB 1IEP)

A Ligand, Bound in Its Pocket

What It Returns

Two Answers About the Pocket

Not a single score, but the pieces you need to reason about a binding site — which residues form it, and what it likely binds.

Binding Residues

A per-residue read on which positions line a binding site — the contact surface, not just a yes/no for the whole protein.

Ligand Types

Which of 17 ligand categories is most likely to bind — nucleotides, cofactors, metals, peptides and more — retrieved from a 2M+ ligand panel.

In Context

The Site, on the Structure

Binding residues are shown against the fold and membrane topology, so the pocket reads as a cluster — not a scattered list of positions.

EXTRACELLULARCYTOPLASMTM1TM2TM3TM4TM5TM6TM7ICL1ECL1ICL2ECL2ICL3ECL3GPPNCPTMBindingDisulfideTM helixpLDDT
Predicted Binding Residues (Diamonds) & Topology · Bovine Rhodopsin (P08100)
Binding Read-Out
Top Retrieved Ligand
Retinal (11-cis)
Small Molecule
Binding Residues
113117122186265296

Illustrative Read-Out for a Well-Characterized Pocket.

AstraBIND: Graph Attention Network for Predicting Ligand Binding Sites
Method and evaluation against public baselines · bioRxiv, 2025
Read →

Recovered in Practice

Real Pockets, Found from Sequence

On well-characterized proteins, AstraBIND recovers the pockets a structural biologist would point to — working from sequence alone.

Recovered Pocket

β2-Adrenergic Receptor

Recovered the orthosteric ligand pocket of the GPCR.

Recovered Pocket

Hemoglobin β-Chain

Located the heme-binding pocket.

Recovered Pocket

Calmodulin

Found all four Ca²⁺-binding EF-hand sites.

2M+
Ligands Screened
17 Curated Categories
250K+
Training Interactions
Protein–Ligand Pairs
< 15 min
Per Protein
Residue-Level Binding Map
Binding-site F1 by ligand class
Nucleotides0.79
Porphyrins (heme)0.74
Cofactors0.73

Strongest where the binding chemistry is well-defined; weighted macro-F1 0.47 across all 17 ligand classes.

Where It's Tested

Binding, Benchmarked Across Families

Binding-site and pocket-triage performance is reported at scale in the open Model Performance whitepapers — across the families where getting the pocket right matters most.

3
Signals Combined
Sequence · Structure · Homology
18,367
Proteins in the Series
Across Five Families
Open
Preprint on bioRxiv
Peer-Facing Methods

Put Astra on Your Targets

Every Astra model runs inside the Orbion platform. Bring a sequence and get the full read-out — structure, function, modifications, binding, and stability.