Allosteric site prediction and transmission of functional residues with atomistic graph analysis

Allostery, the 'second secret of life', that regulates protein activity through molecular binding at a distant site from the orthosteric site, has drawn increasing attention.The universality of allosteric regulation and the great benefits of allosteric drugs make understanding allosteric mechanisms invaluable, yet there are few computational methods to effectively predict allosteric sites. Bond-to-bond (B2B) propensity analysis has successfully identified allosteric sites for a diverse group of proteins with merely the knowledge of the orthosteric sites/ligands using the atomistic energy-weighted protein graphs and outperformed other methods to date.3,4 Crucially, it remains computationally efficient as it scales almost linearly.

However, not all proteins possess clearly defined orthosteric sites which limits their potential as drug targets and renders B2B propensity analysis less effective. Covalent inhibition of protein targets regained popularity over the past two decades and reveal allosteric sites containing inherently reactive amino acids, primarily cysteines, but including lysine, histidine and threonine.5 Targeting non-catalytic cysteines has drawn increased attention from chemical biologists6,7 with chemoproteomic technologies developed to mine chemically accessible cysteines8,9 but limited experimental methods to discern functional cysteines.10,11 Therefore, predicting allosteric sites containing/in proximation to reactive amino acids without the knowledge of the orthosteric sites would enhance cost and time efficiency in the covalent drug discovery process. This PhD focuses on developing atomistic graph anlaysis methods that can identify functional allosteric residues and their paths of allosteric transmission.