Bldg 38A, B2 NCBI Library

 

Speaker: Evan Bolton

 

 

Precomputed 3D similarity of small molecules

 

PubChem precomputes similarity relationships between all small molecules in the PubChem Compound database.  Traditional chemical structure similarity metrics 
are based on the connectivity graph of atoms (so called 2D similarity) and are very efficient in finding chemical structures that are analogs or near-analogs 
of each other.  Presumably, these 2D similar structures with minor variation have similar biological activity and function.  Protein binding pockets interact 
with those small molecules with appropriate 3D shape and 3D orientation of features required for binding.  Chemical structures do not have to have a similar 
connectivity graph to bind to a protein.  As such, many important relationships between small molecules may go unrecognized using 2D similarity metrics alone.  
To help resolve this gap, PubChem has introduced a new similarity metric relating similarity of 3D shape and 3D features of small molecule.  This seminar will 
describe how this new relationship is generated and explore aspects of its utility.