QSAR of caffeines by similarity cluster prediction

Abstract

AbstractA novel QSAR approach based on correlation weighting and alignment over a hypermolecule that mimics the investigated correlational space was performed on a set of 40 caffeines downloaded from the PubChem database. The best models describing log P and LD50 values of this set of caffeine derivatives were validated against the external test set and in a new predictive model by using clusters of similarity.