Premium
From Activity Cliffs to Target‐Specific Scoring Models and Pharmacophore Hypotheses
Author(s) -
Seebeck Birte,
Wagener Markus,
Rarey Matthias
Publication year - 2011
Publication title -
chemmedchem
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.817
H-Index - 100
eISSN - 1860-7187
pISSN - 1860-7179
DOI - 10.1002/cmdc.201100179
Subject(s) - pharmacophore , virtual screening , quantitative structure–activity relationship , computer science , identification (biology) , drug discovery , key (lock) , protein ligand , visualization , set (abstract data type) , cheminformatics , computational biology , data mining , artificial intelligence , machine learning , chemistry , bioinformatics , biology , programming language , botany , computer security , organic chemistry
The role of activity cliffs in drug discovery projects is certainly two‐edged: on the one hand, they often lead to the failure of QSAR modeling techniques; on the other, they are highly valuable for identifying key aspects of SARs. In the presence of activity cliffs the results of purely ligand‐based QSAR approaches often remain puzzling, and the resulting models have limited predictive power. Herein we present a new approach for the i dentification of s tructure‐based a ctivity c liffs (ISAC). It uses the valuable information of activity cliffs in a structure‐based design scenario by analyzing interaction energies of protein–ligand complexes. Using the relative frequency at which a protein atom is involved in activity cliff events, we introduce a novel visualization of hot spots in the active site of a protein. The ISAC approach supports the medicinal chemist in elucidating the key interacting atoms of the binding site and facilitates the development of pharmacophore hypotheses. The hot spot visualization can be applied to small data sets in early project phases as well as in the lead optimization process. Based on the ISAC approach, we developed a method to derive target‐specific scoring functions and pharmacophore constraints, which were validated on independent external data sets in virtual screening experiments. The activity‐cliff‐based approach shows an improved enrichment over the generic empirical scoring function for various protein targets in the validation set.