作者: Nematollah Omidikia , Mohsen Kompany-Zareh
DOI: 10.1016/J.CHEMOLAB.2016.01.007
关键词:
摘要: Abstract Selection of valid and informative descriptors is one the most crucial steps in QSAR studies. In this contribution, type (I, II) errors variable selection proposed as a statistically based method for selecting descriptors. Developed strategy can be considered sophisticated combination jackknifing y-randomization. Type (I) error measure descriptors' importance, it provided using each descriptor. (II) chance correlation calculated Successive projection algorithm Gram–Schmidt orthogonalization were utilized pre-selection methods initial reduction collinear uninformative variables. Selwood data including 31 molecules 53 descriptors, anti-HIV 107 160, Flour 116 1268 research. Model parameters set before after confirm adequacy novel selection.