Premium
Digital Filters for Molecular Interaction Field Descriptors
Author(s) -
Barbosa Euzébio Guimarães,
Ferreira Márcia Miguel Castro
Publication year - 2012
Publication title -
molecular informatics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.481
H-Index - 68
eISSN - 1868-1751
pISSN - 1868-1743
DOI - 10.1002/minf.201000181
Subject(s) - quantitative structure–activity relationship , computer science , relevance (law) , field (mathematics) , feature selection , selection (genetic algorithm) , molecular descriptor , artificial intelligence , machine learning , data mining , protocol (science) , force field (fiction) , filter (signal processing) , model selection , biological system , mathematics , medicine , alternative medicine , pathology , political science , pure mathematics , law , computer vision , biology
Abstract Descriptor properties are often neglected when building 3D‐QSAR models. The relevance of correlation and distribution profiles is tested in terms of the models’ prediction power. A different approach to filter descriptors prior to variable selection is proposed. Additionally, a protocol for molecular interaction field descriptors selection and model validation is presented. The algorithms and protocols presented are quite simple and enable a different and powerful way to create parsimonious interaction field‐based models.