A basic problem in QSAR modeling is how to obtain proper molecular conformations to generate descriptors. In pharmaceutical chemistry, it is ideal if the active conformation can be obtained, otherwise lowest-energy conformation is commonly used. There are different ways to minimize a molecule, but if it can be reasonably minimized to its global lowest-energy conformation instead of a local one is suspectable. Unfortunately this problem is generally neglected by most of the researchers. Considering that different conformations may influence the quality of the subsequent QSAR models, here the conformation searching process is used to ensure that a molecule is minimized to its global lowest-energy conformation, and accordingly QSAR models are evloved. The molecular conformations without searching process, maybe just local lowest-energy, are used as comparison. Three datasets with various structural complexities and flexibilities are investigated. For each dataset, six different lowest-energy conformations are generated and six model populations are established accordingly. The results indicate that different optimization processes can influence the quality of the QSAR models, and global and local lowest-energy conformations have different performance in QSAR modeling. Furthermore, to get the proper global lowest-energy molecular conformation is vital for the subsequent QSAR model development and new query prediction.
The influence of molecular lowest-energy conformation on the quality of the subsequent quantitative structure-activity relationship models
GRAMATICA, PAOLA
2013-01-01
Abstract
A basic problem in QSAR modeling is how to obtain proper molecular conformations to generate descriptors. In pharmaceutical chemistry, it is ideal if the active conformation can be obtained, otherwise lowest-energy conformation is commonly used. There are different ways to minimize a molecule, but if it can be reasonably minimized to its global lowest-energy conformation instead of a local one is suspectable. Unfortunately this problem is generally neglected by most of the researchers. Considering that different conformations may influence the quality of the subsequent QSAR models, here the conformation searching process is used to ensure that a molecule is minimized to its global lowest-energy conformation, and accordingly QSAR models are evloved. The molecular conformations without searching process, maybe just local lowest-energy, are used as comparison. Three datasets with various structural complexities and flexibilities are investigated. For each dataset, six different lowest-energy conformations are generated and six model populations are established accordingly. The results indicate that different optimization processes can influence the quality of the QSAR models, and global and local lowest-energy conformations have different performance in QSAR modeling. Furthermore, to get the proper global lowest-energy molecular conformation is vital for the subsequent QSAR model development and new query prediction.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
