Mentioning HMPA the other day prompts me to talk a bit about the relationship between the two branches of drug-company chemistry: discovery and process. I'm in the first camp - I've done a little of the second, but nothing hard-core. At most companys, there isn't too much crossover, because the two kinds of work are quite different.
Discovery, as you'd figure, involves a lot of different reactions, on different substrates, all to the end of making a lot of different products. Our targets are constantly changing, particularly in the first stages of a new program. If the synthesis of some analog doesn't work out well for us, often the best solution is to drop it and make something else. There's generally something just as good on the list that hasn't been done yet. We like easy, reliable reactions, because those help us generate the widest variety of compounds in the shortest amount of time (and with the least amount of work, come to think of it).
Process chemistry comes into the picture when something has been seriously considered as a clinical candidate, and there's a need to make large, reproducible batches of it. They work on one molecule, and they beat the stuffing out of it. The route that the medicinal chemists used to make the candidate is almost never the best that can be found - at least I've never heard of a case yet where it was. There's always room to use cheaper reagents, higher-yielding reactions (and fewer of them), and solvents that can be dealt with on large scale. And there's the reproducibility issue, too. A synthesis that gives you 90% yields four times out of five and 40% the other time is a disaster. That's an average of 80% yield, but the process chemists would be much, much happier with a lower-yielding route that gives exactly the same yield (with the same degree of purity) every single time.
The process gang will ditch solvents like tetrahydrofuran or (God help you) ether for things like toluene and ethyl acetate. They'll try to get rid of those low-temperature dry-ice cooled reactions, because they'd much rather work in a regular ice bath if possible. All those chromatography steps will be attacked, because they hate running columns on that scale, and who doesn't? Crystallization, precipitations, filtering through a plug of silica gel - anything but running a long column and cutting fractions. If you're a considerate medicinal chemist, you'll have thought about these issues beforehand rather than just throwing the whole problem over the wall when you're done with it. That's why I never use HMPA, because either my reaction can do without it, or we can do without my reaction.
When the quantities involved get serious, some people will step in and see if the entire approach needs to be torn up. There are drugs out there that have had five or ten different routes to them over the years. You don't want to make the whole program depend on finding a new one, but it's worth some work on the side. By this time, the chemistry is moving on to another world in the pilot plant, where the hard-hatted crew worry about issues like starting on the top floor reactor so they can gravity-filter into the room below. When you start thinking about the viscosity of your reaction mixture and the shape of the pipes it's going to be running through, you've moved into the world of chemical engineering.
But meanwhile, people like me are back in the med-chem labs, starting another project on a totally different series of molecules. We're weighing out a hundred milligrams of this and that, trying things out in five-mL flasks to see if they work. It starts again.