AstraBIND Preprint is Published on bioRxiv

We’re excited to announce the release of our latest preprint: AstraBIND — a Graph Attention Network for Predicting Ligand Binding Sites.

AstraBIND bridges sequence and structure-based methods to predict ligand classes and binding residues within minutes. Built on a lightweight GATv2 architecture (0.9 M parameters) and trained on over 250,000 protein–ligand complexes, it achieves state-of-the-art accuracy with rapid inference time.

In benchmarking, AstraBIND reached a weighted macro-F1 of 0.47, performing strongest for nucleotides, porphyrins, and cofactors. Case studies—including p53 and CRFR1—demonstrate robust pocket localization across diverse proteins.

Together with our other Astra models, AstraBIND moves us closer to real-time protein design and validation pipelines.

AstraBIND is already live on the Characterize platform.

Read the preprint here → https://www.biorxiv.org/content/10.1101/2025.11.10.687555v1