G. Tedesco, A. Feriana, A. Pozzan, A. M. Capelli
Glaxo Smith Kline
A fast and readily accessible methodology based on substructural analysis using standard 1024 Daylight and in house developed 3DMill pharmacophoric hashed fingerprints was applied to characterise "7TM-like" and "antibacterial-like" compounds, using different scoring methods. The derived scoring functions were utilised to rank molecules in simulated screening experiments according to their probability of being active respectively as 7TM ligands and antibacterials. Results and performance of the different methods were compared using various criteria, taking into account both the hit rate and the quality of the hits, in terms of their dissimilarity among each other and with respect to the molecules of the training set.