Here’s an update to a post from last year about silicon in drug-like molecules. The Denmark group at Illinois has investigated a range of silicon-containing heterocycles, providing both synthetic routes into the (mostly unknown) structures, and looking at some basic pharmaceutically relevant properties.
There’s a lot of work in this paper on the synthetic procedures, both to prepare these compounds and to investigate their chemistry. What it tells you is that (as you would have figured) that C-to-Si is a nontrivial switch in the fume hood, because the chemistries involved can be quite different. For instance, that first core on the list does N-alkylation and N-acylation reactions just fine, but when you try to do metal-catalyzed N-C couplings on it, things go to pieces (protodesilylation and other decomposition reactions). Protodesilylation in general was something that had to be looked out for, with attention paid to solvent choice and temperature in the reaction conditions.
Another theme that comes up several times is that the nitrogens, when exposed to base, can be significantly more prone to oxidative side reactions than expected (even fairly simple reductive amination reactions had to be performed under inert atmosphere, although they were successful then). Metalation reactions were also tricky. Some of them worked just fine and reacted as expected with a range of electrophiles, while others would work with some partners but not others. But some of them just fell apart immediately on attempts to form the anions (such as C2 of the second compound on the list shown), and the only way to find that out was to try them.
Overall, the group was able to prepare an impressive range of functionalized derivatives of these cores, almost all of which are shots into unknown territory. How, then, do such compounds behave? Their cLogP values are almost invariably higher than their carbon analogs, but as usual when you move into unusual structure space, it would be interesting to have some experimental values to make sure that that some of this just isn’t a problem with the calculations.
Comparing the Si and C matched pairs in assays, there seems to be no difference in P-gp transporter behavior in vitro, for starters. There was a slight trend towards more instability against CYP enzymes, but this wasn’t universal (and was also species-dependent, with the rat enzymes being more vigorous. (That’s often the case; rats have a reputation for having more strongly oxidizing liver enzymes, which I’ve always attributed to their rather broad definitions of suitable food). Inhibition of these enzymes didn’t show much of a trend one way or another.
The second compound on the list above and its carbon analog were compared in rat PK experiments as well (oral and i.v.) Intravenous behavior was fairly similar – the Si compound had slightly higher clearance and slightly higher volume of distribution. As for p.o. dosing, the Si compound is significantly more bioavailable, although it has a lower Cmax. Overall AUC was pretty much the same, though. And even though it had greater instability to the rat enzymes, it had a twofold great half-life in vivo, which makes you think plasma protein binding. So this is only one point, but taken with the other Si compounds in the literature, it seems as if they fall inside the normal range of variation that you see with all-carbon compounds – there’s nothing intrinsically weird about them from a pharmacokinetic/metabolic standpoint.
That brings up what I mentioned in my past last year, though: perhaps it would be better for silicon-containing drugs if there were something unusual about them. Admittedly, that could also be “unusually bad”, but overall, it’s harder to make the case for moving to silicon if the effects of doing so are (a) not huge and (b) not all that predictable. You’re already in that zone with carbon analogs, most likely, so why bother? The paper itself has this to say:
This lack of success in the pharmaceutical industry may be attributed to two key factors: (1) an absence of general and accessible synthetic methods for the construction of appropriately functionalized silicon-containing molecules and (2) ineffective approaches to the utilization of silicon, of which the “carbon/silicon switch” is the most common.
I think that second point is what we’re talking about. “Ineffective approaches to the utilization of silicon”, from another angle, means “lack of a good reason to use it at all”. If some effective uses for it can be found – and they may well be out there, who knows? – then things will change. But not until then. If the late-stage silicon switch doesn’t necessarily get you anything, it’s true that you’re going to have to look earlier (good activity in a silicon-containing compound that isn’t replicated by its carbon analog), and this work is an attempt to provide a host of new Si-containing chemical matter towards that end.
What would go a long way to answering the overall question would be if someone were to produce a library of (say) ten or twenty thousand diverse silicon-containing drug-like structures and their exact carbon analogs, and do some high-throughput screening campaigns (variously targeted, cellular, and phenotypic) to see how they behave. But no one’s going to go to that trouble just yet, for just the reasons described (an absence of ways to make such things and an absence of compelling reasons to make them). The chicken and egg question remains.