There are different schools of thought regarding the preparation of ligands for either ligand-based drug design (LBDD) or structure-based drug design (SBDD).

In LBDD, molecular descriptors of the ligands are used to construct QSAR models to predict biological activity. Alternatively, knowledge of key features of ligands for binding to a receptor of interest is employed to construct pharmacophore models.

In SBDD, ligands are screened for favorable binding to a given receptor by ligand-receptor docking.

With either LBDD or SBDD, some investigators prefer to start with a rigorous QM-based energy minimization of ligands using software such as Gaussian or GAMESS. Because QM-based minimization is computationally intensive and requires relatively long times, others prefer to start with relatively rapid minimization using software based on force fields such as AMBER or MMFF94.

The optimal conformation of the ligand in the binding site of the receptor is not necessarily the energy-minimized ground state, which is one reason for allowing the ligand to be flexible during the docking procedure. In addition, because docking generally cannot completely sample the conformational space of the ligand, some investigators use specialized software to generate a library of conformers for use in docking.

The ligand structure should be optimized, and partial atomic charges assigned, as recommended for the given force field; any quantum-mechanical software can be used for this purpose.

Rudy J Richardson thanks for your detail answer.