Scoring function components optimized in this work.

The components are split into three classes, depending on their sign: negative (attractive), positive (repulsive). The Coulomb potential is attractive or repulsive depending on the product of charges. Each component XXX-x is labeled with an amino acid, followed by a dash, then a nucleic acid moiety, which can be one of: phosphate group, P; major groove, M; and minor groove, m. The |CCP| component is the vector length of the Chemical Context Profile, which is proportional to the loss of accessible surface area upon binding. Pair-wise interactions selected by the Forward Sequential Feature Selection methods are in the form XXX-x.