Of the physical properties that make up the "Rule of Five" (and similar schemes), the one that I think is easiest to breach is molecular weight. I'm not saying that it's a good idea to breeze past 500 daltons with a song on your lips - you should always realize that you're probably asking for trouble up there. But trouble seems to follow a bit less often than it does with, say, a high logP. For a high-value target, I think it's certainly worth pursuing if that's really where you have to go.
Here's a whopper of a molecule, for example, an inhibitor of PTP-MEG2 (also known as PTPN9). That's an unusual phosphatase involved in hepatic insulin signaling, and it's already been shown that knocking it down seems to be beneficial in diabetic rodent models. But, like another longtime diabetes target in this space (PTP1B), it's not easy to get a decent inhibitor. Phosphatases are tricky. Their active sites are very polar (as you'd imagine, having to work with phosphate anions all day), and there aren't all that many phosphatase subtypes as you'd expect, given the amount of such work there is to do. That leads to worries about selectivity, even should you find a molecule that seems to work.
So if you can't find a decent inhibitor, how about an indecent one? That's the first reaction on seeing the structure at left. You can certainly see its resin-bound peptidomimetic library roots. I only wish the authors had found a way to incorporate a chlorine atom somewhere; it would have been one of the rare compounds that runs the table on the halogens. As it is, this floating island weighs 1084, well beyond what anyone could consider reasonable for a drug candidate.
It's selective, naturally. Something this size is making so many interactions that its chances of fitting in a lot of different places is quite small. (There's a crystal structure, which doesn't appear to be showing up in the PDB as yet). It's got a Ki of 34 nanomolar against its target, and about 600 for PTP-TC and PTP1B. All the other protein tyrosine phosphatases are dead, and I'd be very surprised if it hits something from another class. But selectivity isn't the hurdle for these leviathans - it's pharmacokinetics. And here we have a surprise.
First off, the compound shows good activity in mouse heptatocytes, and in other insulin-sensitive cell lines. That's quite interesting, since PTP-MEG2 is surely intracellular - so what part of the cell membrane is letting Godzilla through the turnstiles? The authors moved on to i.p. injection in mice, and found that at at 20 mpk level the compound hit a Cmax of 4.5 micromolar (pretty respectable, considering that molecular weight), and had a half-life of 1.8 hours. That's short, but not as short as one might have feared. Multiday treatment of mice showed just the sorts of on-target effects that one might have predicted: inhibition of hepatic gluconeogenesis, enhanced glucose clearance and insulin sensitivity. That's just the sort of profile you'd want for a Type II diabetes drug, and with a bit of work on the half-life, you might have one here as an injectable. I don't hold out much hope for oral activity with a molecule like this, but it's impressive that it gets this far, and it provides some real proof-of-concept for PTP-MEG2 as a drug target.
So in case anyone's wondering whether I can say anything kind about tool compounds, or about academic drug discovery (this paper's from Indiana U), well, here's your evidence. I don't know whether the authors were brave or just foolhardy to consider these structures, but they've latched onto something worthwhile.