| Abstrakt: |
Abstract To estimate how sophisticated should an empirical scoring function be to ensure successful docking, scoring and virtual screening a new scoring function NScore (naive score) has been developed and tested. NScore is an extremely simple function and has the minimum possible number of parameters; nevertheless, it allows all the main effects determining the ligand–protein interaction to be taken into account. The fundamental difference of NScore from the currently used empirical functions is that all its parameters are selected on the basis of general physical considerations, without any adjustment or training with the use of experimental data on ligand–protein interaction. The results of docking and scoring with the use of NScore in an independent test sets of proteins and ligands have proved to be as good as those yielded by the ICM, GOLD, and Glide software packages, which use sophisticated empirical scoring functions. With respect to some parameters, the results of docking with the use of NScore are even better than those obtained using other functions. Since no training set is used in the development of NScore, this scoring function is indeed versatile in that it does not depend on the specific goal or target. We have performed virtual screening for ten targets and obtained results almost as good as those yielded by the Glide and better than GOLD and DOCK software. Figure Average percent of known actives found vs percent of the ranked database screened (x axis) for NScore (NScore, black)- an extremely simple function where all parameters are selected on the basis of general physical considerations, without any adjustment or training with the use of experimental data, Glide XP (XP, red), Glide SP (SP, green), DOCK (DOCK, blue), GOLD GoldScore1x (gold1x, cyan), and GOLD ChemScore1x (chem1x, magenta). Grey lines (rand show results expected by chance. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink