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

Area of Science:

• Biochemistry
• Pharmacology
• Computational Biology

Background:

• Membrane proteins are crucial therapeutic targets but lack comprehensive quantitative ligand-binding data.
• Existing datasets often lack standardization, hindering comparative analysis and drug discovery efforts.

Purpose of the Study:

• To establish Binder2030, a curated dataset of small-molecule ligands and transmembrane protein interactions.
• To provide standardized dissociation constant (Kd) measurements for comparative analysis across diverse target classes.
• To facilitate downstream applications in drug discovery and structure-based modeling.

Main Methods:

• Affinity selection-mass spectrometry (ASMS) using Binder Selection Technology (BST) on membrane fractions.
• Standardized Kd measurements, chemical identifier curation, and target annotations.
• Development of a PubChem-overlap subset with activity annotations for external validation.

Main Results:

• Binder2030 dataset includes 3,384 small-molecule ligands targeting approximately 400 transmembrane proteins (GPCRs, SLCs, ion channels).
• Standardized Kd values enable analysis of affinity distributions and chemical space across target classes.
• Demonstrated successful integration with structure-based modeling, comparing predicted potencies with experimental affinities.

Conclusions:

• Binder2030 significantly expands quantitative ligand-binding data for membrane proteins.
• The dataset and associated tools support comparative analysis and accelerate the discovery of novel therapeutics.
• Facilitates structure-based drug design by linking experimental affinities with computational predictions.