Here's a paper from some folks at GlaxoSmithKline on what kinds of rings seem to have the best chances as parts of a drug structure. They're looking at replacements for plain old aryl rings, of which there are often too many. Pulling data out of the GSK corporate collection, they find that the most common heteroaromatic rings are pyridine, pyrazole, and pyrimidine - together, those are about half the data set. (The least common, in case you're wondering, are 1,3,5-triazine, 1,3,4-oxadiazole, and tetrazole). In marketed drugs, though, pyridine is more of a clear winner, and both pyrrole and imidazole make the top of the charts as well.
When they checked the aqueous solubility of all these compounds, the 1,2,4-triazoles came out on top, and the 1,3,5-triazines were at the bottom, which sounds about right. Other soluble heterocycles included 1,3,4-oxadizole and pyridazine, and other bricks were thiazole and thiophene (not that that last one really counts as a heterocycle in my book). Update: I've revised my thoughts on that! Now, you might look at these and say "Sure, and you could have saved yourself the trouble by just looking at the logD values - don't they line up?" They do, for the most part, but it turns out that the triazines are unusually bad for their logDs, while the five-membered rings with adjacent nitrogens (all of 'em) were unusually good.
The next thing the team looked at was binding to human serum albumin. The 1,3,4-thiadiazoles emerged as the losers here, with by far the most protein binding, followed by thiazoles and 1,2,4-oxadiazoles. Imidazoles had the least, by a good margin, followed by pyrazine and pyridazine. Those last two were better than expected compared to their logD values.
And the last big category was CYP450 inhibition. Here, thiophene, tetrazole, and 1,2,3-triazole were the bad guys, and pyridazine, 1,3,4-thiadizole, and pyrazine (and a few others) were relatively clean. The people at AstraZeneca have published a similar analysis, and the two data sets agree pretty well, with the exception of oxazole and tetrazole. The AZ oxazoles all had open positions next to the ring nitrogen, which seems to have opened them up to metabolism, but the difference in tetrazoles (AZ good, GSK bad) is harder to explain.
The take-home? Pyridazine, pyrazine, imidazole and pyrazole look like the winners from an overall "developability" score. Thiophene brings up the rear, but since I still think that one shouldn't count update (
it's a benzene in disguise), the ones to worry about are then thiazole, 1,2,3-triazole, and tetrazole (that last one with an asterisk, due to the CYP data discrepancy).
The paper tries to do the same analysis with heteroaliphatic rings, but the authors admit that they had a much smaller data set to work with, so the conclusions aren't as strong. There was also a higher correlation with plain ol' logD values across all three categories (not as many surprises). The winners turned out to be piperidine NH and morpholine N-alkyl, with imidazoline and piperidine N-alkyl right behind. The losers? Piperidine N-sulfonamide, followed by pyrrolidine N-sulfonamide, and then 1,3-thiazolidine. (Sulfonamides continue to live up - or down - to their reputation as Bad News).
There are, naturally, limitations to this sort of thing. Ceteris paribus is a mighty difficult state of affairs to achieve in medicinal chemistry, and other factors can rearrange things quickly. But if you're just starting out in an SAR series, it sounds like you might wand to give the pyrazines and pyridazines a look.