As a drug discovery project goes along, different labs tend to claim different parts of the molecule to work on. They run all sorts of variations within their territory, usually keeping the rest of the molecule at some sort of agreed-on default setting or two. Likely as not, they'll find something along the way that makes things a lot better (more potent, longer-lasting in the blood, etc.)

The natural thing to do is to combine these things, to make what I've long called a greatest-hits molecule. "Let's put that acyl group that Jim likes on there, and put best of the N-aryls from Sue's lab, and over on the side chain we'll have that solubilizing group that works so well on Wei's molecule. . .can't miss!"

Actually, these things miss about as often as they hit. Rarely have I seen a project where you can mix-and-match with confidence. You have to try these combinations, but after you've started to fill out the matix, you find that your compounds act more like this: "Well, the acyl group is good, as long as you don't have a heteroaryl group over here, but if you do then you can get away with the chloro on this position, but not if you have the amine side chain, except when there's an alpha-methyl. . ."

What's going on is that your molecules probably don't have just one way of fitting into their binding pocket in the target protein. They might have two modes; they might have twelve. There's no way to be sure, and I say that with no intention of offending the molecular modelers and their computer simulations. (But hey, if the shoe can be docked onto your foot in a low-energy conformation, wear it.) Many of these binding modes are going to have mutually incompatible features, and you can make your head vibrate trying to reconcile them into a single coherent picture.