SandboxAQ Releases SAIR, a 5.2 Million-Structure Open Dataset for Drug Discovery AI

SandboxAQ released SAIR (Structurally Augmented IC50 Repository), a free public dataset of roughly 5.2 million synthetic protein-ligand 3D structures spanning more than 1 million distinct protein-ligand systems. The data was generated using NVIDIA DGX Cloud computing power and the Boltz-1x co-folding model, and is meant to help researchers train and benchmark AI models that predict binding affinity, how strongly a drug molecule attaches to its target protein, a key step in early-stage drug discovery. SandboxAQ is a Parkway Venture Capital portfolio company.

Read the full article: SandboxAQ's dataset announcement