Premium
3D‐QSAR Applied to the Quantitative Prediction of Penicillin G Amidase Selectivity
Author(s) -
Braiuca Paolo,
Boscarol Luca,
Ebert Cynthia,
Linda Paolo,
Gardossi Lucia
Publication year - 2006
Publication title -
advanced synthesis and catalysis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.541
H-Index - 155
eISSN - 1615-4169
pISSN - 1615-4150
DOI - 10.1002/adsc.200505346
Subject(s) - quantitative structure–activity relationship , chemistry , grind , molecular descriptor , penicillin amidase , selectivity , biological system , enzyme kinetics , substrate (aquarium) , computational chemistry , combinatorial chemistry , penicillin , stereochemistry , active site , enzyme , organic chemistry , catalysis , biochemistry , antibiotics , mechanical engineering , grinding , oceanography , geology , engineering , biology
A new approach for predicting the selectivity of penicillin G amidase (PGA) – expressed as k cat /K M – is here described. Regression models were constructed correlating the experimentally determined k cat /K M of a limited number of substrates to molecular descriptors calculated by using methods generally employed in drug discovery for quantitative structure‐activity relationship (3D‐QSAR methods). Two different methods for the calculation of molecular descriptors have been tested, namely GRIND and Volsurf. The real predictions, made on molecules not used for constructing the models, had an accuracy sufficient for being useful in the experimental practice. Both approaches led to models able to predict substrate selectivity even without modelling the enzyme‐substrate complex, whereas the prediction of enantioselectivity was feasible only by combining the GRIND approach with the conformational analysis of the substrates inside the enzyme's active site. The present approach represents an actual alternative to screening procedures since it allows one to develop a whole predicting model in a few hours, once a small set of experimental data is made available.