Hi Ariya,

After docking, count the number of hydrogen bond between ligand and receptor. Ligands with better scoring energy and 1 or 2 hydrogen bonds should show better activity.

In addition to lipinski rules, there are other molecular descriptors that allow you to rank compounds. As you want to decrease number of compounds you can clustering them by chemical similarity and select a representative structure of these groups. There are several softwares that allow performing these actions, for example (MOE, DataWarrior, etc.). I recommend MOE because it has a wide variety of molecular descriptors and is easy for use.

You must take into account several things:

-the rules of lipinski, like the other molecular descriptors are an approximate measure of possible pharmacokinetic properties of these compounds. Then, if you are interested in the design of drugs with better pharmacokinetic properties, potent or selective, eliminating some of these compounds because they do not comply with a particular rule will reduce the chances of identifying the best hit, that is, if you eliminate a compound that its molecular weight (MW) was 501 Da, from substitution of some groups, you can obtain a compound of lower MW and with better properties than the previous one.

-similar occurs with grouping by chemical similarity. In this way the number of compounds decreases considerably, but you lose a lot of information.

You can still filter the compounds based on other parameters like Ghose filter, QEM, Toxicity, Target-mediated, etc.