Yesterday's post set off a discussion of the 1990s combichem boom in the comments. I joined the industry before that took off, and watched it with interest.

For those outside the field, combinatorial chemistry was (is, I guess) the semi-automated generation of large numbers of diverse organic compounds. The basic idea was that you'd start with, say, building block A, which would react with a big library of reactant partners B1. . .Bzillion. The resulting compounds would then be reacted with another big set of coupling partners, C1. . .Cmonstrous, and this might be designed to take place out at the end of the B part, or on another part of the A region, etc. There were many pool-split-mix methods worked out to generate the maximum number of different compounds. Various strategies generated either individual compounds or mixtures of different ones and all sorts of techniques were developed to make it all happen in a less labor-intensive fashion. These included bonding the starting materials (or the reagents) onto solid resin bead supports until the end of the synthesis, the better to move things around, along with ingenious schemes for tagging and identifying what was ending up in which vials.

The idea was that we'd generate lots (lots) more compounds for random screening than we'd ever have before. And for a while, it looked like the companies that did this the first with the most were going to have the drop on everyone else. It stood to reason - many of our high-throughput screens didn't generate anything useful to start working on, so if technology now allowed you to brute-force your way into getting things to hit, well, you'd be crazy not to.

A frenzy ensued. People that no one had heard of were suddenly in demand as consultants and invited speakers at conferences. Whole companies were started to make and sell combinatorial libraries of compounds - a couple of them are even still in business, although the road has been pretty jumpy. Larger companies started in-house efforts, some of them rather lavish. Some people talked about traditional medicinal chemistry receding to a specialty, as the mighty compound factories came on line (more than one person tried to sell me on this idea personally).

But as time went on, and the piles of combichem stuff made it into the screening collections, people began to note with unease that, well, not so many lead compounds were coming out. In fact, it eventually became clear that the hit rate for most combichem stuff was lower than for the general old-fashioned screening collections. That went double for the combi libraries from the first part of the boom, many of which are now regarded as basically worthless.

What happened? Well, the techniques that generated larger mixtures of compounds were trouble from the start, because it's hard enough to screen individual compounds well. But even single-compound collections had their problems. A larger difficulty was that the chemistry that could be used under the more highly automated combichem protocols was limited. Many useful reactions were bypassed because there was no good way to do them on solid supports with minimal purification afterwards. There sure were an awful lot of amides, ureas, and sulfonamides produced, I can tell you. Not that there's anything wrong with these groups, but when you start to have multiple instances of them in the same molecule, you can veer off into undesirable territory.

Overall, as has been realized, the chemical diversity offered by combichem's early years was largely spurious. People went out and did the stuff that was easiest to do, with what was on hand, and that translated to a much spottier coverage of chemical space than was first realized. Combichem itself survives, but compared to the mid-1990s it's a backwater.

But there's still a place for it. People have been steadily introducing a greater variety of chemistry into it, and everyone's now more aware of how hard it is to make truly diverse compound collections. Once the hot air hissed out of it, combichem was revealed as what it really had been all along: a tool. One of many, to be used as appropriate.

Update: Here's a take on the field from the inside, from Org Prep Daily.