Our scoring function consisted of a simple empirical scoring function and a pharmacophore-based scoring function to reduce the number of false positives. The energy function can be dissected into the following terms:

E_tot = E_bind + E_pharma + E_ligpre

where E_bind is the empirical binding energy used during the molecular docking ; E_pharma is the energy of binding-site pharmacophores; E_ligpre is a penalty value if the ligand unsatisfied the ligand preferences. E_pharma and E_ligpre were used to improve the number of true positives by discriminating active compounds from hundreds of thousands of non-active compounds. The empirical binding energy (E_bind) is given as

E_bind = E_inter + E_intra + E_penal

where E_inter and E_intra are the intermolecular and intramolecular energy, respectively, E_penal is a large penalty value if the ligand is out of range of the search box. In this paper, E_penal is set to 10000. The intermolecular energy is defined as

r_ij^B_ij is the distance between the atoms i and j with the interaction type B_ij forming by the pair-wise heavy atoms between ligands and proteins; B_ij is eithera hydrogen bond or a steric state; q_i and q_j are the formal charges and 332.0 is a factor that converts the electrostatic energy into kilocalories per mole. The lig and pro denote the numbers of the heavy atoms in the ligand and receptor, respectively. is a simple atomic pairwise potential function (Figure 1.) In this atomic pair-wise model, the interactive types are only hydrogen binding and steric potential which have the same function form but with different parameters, V1, . . . , V6 (defined in Figure 1.) In this model, the atom is divided into four different atom types: donor, acceptor, both, and nonplar. The hydrogen binding can be formed by the following pair atom types: donor-acceptor (or acceptor-donor), donor-both (or both-donor), acceptor-both (or both-acceptor), and both-both. Other pair-atom combinations are to form the steric state.

F(r_ij^B_ij) is defined as above except the value is set to 1000 when r_ij^B_ij < 2.0 angstroms and dihedis the number of rotatable bonds. We followed the work of Gehlhaar et al. (1995) to set the values of A, m, and θ₀. For the sp³-sp³ bond A, m, and θ₀ are set to 3.0, 3, and π; and A = 1.5, m = 6, and θ₀ = 0 for the sp³-sp² bond.

The pharmacophore-based interaction (E_pharma) between the ligand and the protein is calculated by summing up all hot-spot atoms:

where w_j is the pharmacophore weight of the hot-spot atom j, F(r_ij^B_ij) is defined as above, lig is number of the heavy atoms in the ligand, and hs is the number of hot-spot atoms in the receptor. The value of f(w_j, B_ij) is w_j or 0. f(w_j, B_ij) is w_j if the interaction type (B_ij) equals to the type of hot spots found on the target receptor. The ligand preferences include electrostatic (i.e. the number ofelectrostatic atoms) and hydrophilic characteristic (i.e. the atom numbers of hydrogen donor and acceptor). The E_ligpre is a penalty value for a ligand which is unable to satisfy the ligand preferences and is defined as

E_ligpre = WP_elec + WP_hb

where WP_elec and WP_hb are the penalties for the electrostatic and hydrophilic preferences, respectively.