Should I ever be invited to give one of those "Alternative Careers" talks at one of my alma maters, I plan to spend a fair amount of time explaining how the drug development process works. Even for people with no intention of working in the field, it seems like something you ought to understand if you believe that your research ultimately contributes to it.
Here's how academic biologists think drugs are made: Pharma researchers read a PNAS paper about a new kinase and throw it into a big machine with millions of compounds, and a drug comes out. They give it to mice or monkeys, and if it doesn't kill them, give it to people, give other people a placebo, and if it works better than the placebo, it's time to start making the TV commercials.
Here's how it typically works.
Target identification and assays: This is where the majority of molecular/cell biologists are employed, it's what many biotechs do almost entirely, and it employs the people whose academic training prepares them the least for industry. There are different ways to pick the target of interest, largely depending on the company. Some do rely on those PNAS papers, with the biologists replicating the results to find the ones that aren't, y'know, wrong, and trying them out in vivo when the original authors didn't do that. Large companies have have their own basic research and new startups often focus on the founder's PNAS paper from his academic lab. In any case, the goal is to find a target where up- or down-regulation might plausibly be therapeutic, and then to develop a high-throughput assay for that up- or down-regulation.
Screening and chemistry: Given an assay, someone then runs lots of compounds through it. Large pharmas have huge collections that they've accumulated over the last century; small startups might outsource this part. The best candidates serve as starting points for optimization by the synthetic chemists, a process that's a curious mix of science and art. Those new compounds get screened themselves, usually in a better but more labor-intensive assay. Meanwhile, a bunch of supporting cast members chime in with advice from their respective specialties. Lawyers start filing patents around now.
Formulation, scale-up and PK/PD: Formulation is definitely something I'd never imagined before getting involved in drug development. The top candidates coming out of the in-vitro assays have to be turned into something that can be administered to animals, so chemists have to tweak their salts and solvents. When they get the stuff to dissolve, it's given to mice or rats, and pharmacologists study where the drug goes (pharmacokinetics, PK) and biologists look to see if it's doing what it's supposed to do (pharmacodynamics, PD). This is an iterative process, with the formulation people trying to improve the PK numbers. Meanwhile, another group of specialized chemists figure out how to make kilograms of the compound instead of mere grams, and pharmacologists start looking into how the stuff is metabolized.
Toxicology: This is where the compounds are given to rats, dogs or monkeys to identify problems with toxicity, and also to check the PK/PD in primates. This is usually outsourced to specialist testing companies, but still employs primarily pathologists and veterinarians to evaluate the results.
If a compound survives all this, you have a potential drug! For small biotechs, this is the part where they get paid. In a large pharma, it might be where the process starts, when your company in-licenses the drug from the biotech. (After a team of chemists, biologists, pathologists and lawyers in your company do the due diligence to decide whether to buy it.) This seems like a logical point to break...
Note: for biologicals (antibodies, peptides, RNA therapies), the route is a bit different than for "small molecules". Here, a target is also defined, but the synthesis, screening and scale-up stages obviously employ many more biologists instead of chemists. And given the species-specificity of these therapies, the appropriate counterparts to the animal studies mentioned are frequently murky. This sort of development is going to be increasingly important in the future, though, especially if siRNA ever gets working.