Due to their low binding affinities, detecting small-molecule fragments bound to protein structures from crystallographic datasets has been a challenge. Here, we report a trove of 65 new fragment hits for PTP1B, an "undruggable" therapeutic target enzyme for diabetes and cancer. These structures were obtained from computational analysis of data from a large crystallographic screen, demonstrating the power of this approach to elucidate many (∼50% more) "hidden" ligand-bound states of proteins. Our new structures include a fragment hit found in a novel binding site in PTP1B with a unique location relative to the active site, one that links adjacent allosteric sites, and, perhaps most strikingly, a fragment that induces long-range allosteric protein conformational responses. Altogether, our research highlights the utility of computational analysis of crystallographic data, makes publicly available dozens of new ligand-bound structures of a high-value drug target, and identifies novel aspects of ligandability and allostery in PTP1B.
Keyphrases
- small molecule
- protein protein
- high resolution
- electronic health record
- big data
- binding protein
- genome wide
- gene expression
- papillary thyroid
- amino acid
- adipose tissue
- single molecule
- rna seq
- mass spectrometry
- young adults
- transcription factor
- single cell
- artificial intelligence
- skeletal muscle
- adverse drug
- dna binding