Estimation of 2D autocorrelation descriptors and 2D Monte Carlo descriptors as a tool to build up predictive models for acetylcholinesterase (AChE) inhibitory activity

• Published in: Chemometrics and intelligent laboratory systems 2019-01, Vol.184, p.14-21
• Main Authors: Velázquez-Libera, José L., Caballero, Julio, Toropova, Alla P., Toropov, Andrey A.
• Format: Article
• Language: English
• Subjects: 2D autocorrelation space; AChE inhibitory activities; CORAL software; Monte Carlo method; QSAR
• ISSN: 0169-7439; 1873-3239
• DOI: 10.1016/j.chemolab.2018.11.008

Two approaches of building up predictive models for acetylcholinesterase inhibitory activity of a broad dataset of 403 compounds, 2D Autocorrelation descriptors and 2D Monte Carlo descriptors, were compared. In a first approach, the molecular information has been encoded in 2D autocorrelation descriptors, obtained from different weighting schemes. Quantitative structure-activity relationships were built by multiple linear regression combined with evolutionary search. All 2D autocorrelation models were predictive according to internal and external validation experiments, including more than 30 descriptors. The 2D Autocorrelation space brings a major influence of Sanderson electronegativity weighted terms for three different splits. The Monte Carlo descriptors were calculated with the CORAL software (http://www.insilico.eu/coral). Advantages and disadvantages of the above approaches are discussed. •A broad dataset of 403 AChE inhibitors is compiled.•QSAR model considering 2D autocorrelation descriptors was developed.•QSAR model considering 2D Monte Carlo descriptors was performed.•Models were validated according to classical QSAR tests.