Binder2030: a quantitative membrane proteome binding dataset enabling AI-driven drug discovery.

Membrane proteins represent more than half of therapeutic targets but remain underrepresented in quantitative ligand-binding datasets. Here we report Binder2030, a curated affinity selection-mass spectrometry (ASMS) dataset comprising 3,384 small-molecule ligands across approximately 400 transmembrane proteins, including G protein-coupled receptors (GPCRs), solute carrier (SLC) transporters, and ion channels. Using Binder Selection Technology (BST) applied to membrane fractions, Binder2030 provides standardized dissociation constant (Kd) measurements with curated chemical identifiers and target annotations, enabling comparative analysis of affinity distributions and chemical space across target classes. To support external anchoring, we release a PubChem-overlap subset with matched activity annotations and Binder2030 Kd values. Finally, we demonstrate downstream integration in a structure-based modeling workflow by comparing Boltz-2 predicted potencies with experimental affinities for a GlyT-1 ligand set.