抄録
Background: The term druggable genome was coined in 2002 for the estimate of 10% of the human proteome likely to bind small molecules with lead-like chemical properties and sufficient binding affinity for activity modulation. A number of databases now provide protein target lists based largely on chemistry-to-protein mappings extracted from the literature. The IUPHAR/BPS Guide to PHARMACOLOGY (GtoPdb) contains expert curated target sets at a high stringency and updated approximately bi-monthly [1]. This work introduces the statistics of GtoPdb proteome coverage and compares this with other established sources.
Methods: The first step was to review the statistics of GtoPdb for proteins with ligand interactions and the subset where those interactions have quantitative binding data. We then turned to UniProt where we have submitted cross-references. We also used chemistry cross-reference in UniProt to extract target counts for other sources. The intersects and differences were analysed by Venn diagrams, Gene Ontology distributions and other types of cross-reference counts.
Results: GtoPdb includes 1684 human protein targets with ligand interactions with 1489 having quantitative binding data for 6716 ligands. Many of these could provide plausible starting points for drug research and our data show 2089 ligands have clinical data. However, only 313 proteins are the primary targets of 1233 approved drugs, reflecting the validation bottleneck for new targets. We retrieved the following counts for human target counts in other UniProt sources for DrugBank 2376, ChEMBL 3310 and BindingDB 2474. The intersect between all four was only 517. Gene Ontology functional categories, target class content, 3D structure, pathways, disease genes were analysed for the GtoPdb and the consensus sets
Conclusion: This work presents a concise data-supported druggable proteome snapshot from GtoPdb and comparative coverage for three other resources. These four show surprisingly divergent curatorial selectivity. However, the consensus is increasing slowly and the union covers nearly 20% of the proteome. The utility of data-supported druggable proteome lists is high, particularly when genetic data points to new targets likely to have at least some chemical starting points for validation experiments.
References: [1]] Harding et al. (2018). Nucl. Acids Res. 45 (Database Issue)
