As a drug discovery project moves along, we synthesize lots of new compounds, test them, and pick the best ones to make in large quantities. Simple, eh? Try your hand, then, at some of these questions, all of which have come up in the course of my career so far:
1. If you're running an experiment in vivo, and your control compound (from a competitor) is a highly active, hard-to-beat standard - how do you interpret your results when you know that this compound has made it to market and is no great shakes in human patients?
2. What do you do when you have to make a large batch of some compound for advanced pre-clinical work, and there's only one person in the whole department who can really get the crucial reaction to work? Do you tell people that you have a good large-scale route, or not?
3. How about a bit earlier in the game - how do you deal with it when you have a high-yielding, clean route to a key intermediate that lots of your people are using, but it uses a reaction that you know, for a fact, that the scale-up group won't touch later on?
4. How do you handle things when your primary biological assay keeps acting up - by factors of five to ten? Do you normalize the numbers to a standard each time and hope for the best, or do you start to doubt the usefulness of the whole assay?
5. For bonus points, how do you tell which numbers you've been getting are closer to the truth - the ones that say your compounds are really active, or the ones that say that they stink?
6. How do you interpret things when the in vivo assay tells you that your compounds have wonderfully long durations of action, but the blood levels tell you that they completely disappeared from circulation long before?
7. What does it mean when your best compound is intolerant of even slight structural changes? Do you just run with it (after all, you only need one compound, right?) Or do you hammer away trying to find something that can be safely modified in order to have a back-up?
Are there right answers? Well, presumably. I know what answers I'd give to some of these, but I make no guarantees that they're the right ones. . .
1. qetzal on August 1, 2005 8:43 PM writes...
I love the overall theme of this post: how do you decide which data to ignore!
It's amazing how often things come down to that. There's almost never enought time and resources to resolve all the discrepancies. Frequently, you just have to cross your fingers and move ahead, hoping.
In my experience, this is also what causes "zombie projects" - projects that just won't die. Someone can almost always come up with a reason why they might still work.
Permalink to Comment2. John Novak on August 2, 2005 1:00 AM writes...
I bet I could put together analogous questions for most if not all of these in the defense electronics industry.
I bet I could think of people who habitually give (what I consider to be) the wrong answers, too.
Permalink to Comment3. Peter Ellis on August 2, 2005 4:07 AM writes...
No idea about most, but for number 6 I'd hazard a guess that there's an active metabolite that isn't being picked up when you look at blood levels. Either that, or could there be a possibility that the drug is being sequestered out of circulation somewhere it's still able to be active?
Both interesting avenues for further investigation, at the least.
Permalink to Comment4. bamh1d on August 2, 2005 7:07 AM writes...
Pardon me for saying so, but question #4 is just the typical complaint of the medicinal chemist, although it is a question that is easily answered with a little database work and some basic applied statistics. Blaming the assay is relatively easy and cheap, if you make no effort to understand how an assay is designed, developed, and validated, and its performance over time. The usefulness of the assay can easily be assessed by looking at the assay performance with reference standards over time. If the control charts show a range of 5-10 fold with standards, then the assay is crap and should be rebuilt. On the other hand, if you're seeing 5-10 fold variation for your compound and this is statistically greater than the variance for the control charts over the same period of time, then maybe you want to take a look at the solubility or stability of the stuff you're synthesizing? This thought usually occurs only well after mud has been launched at the assay.
Permalink to Comment5. Derek Lowe on August 2, 2005 7:44 AM writes...
Bahm, that's the way to do it, all right. Problem is, many times people want to hold on to assays where the standards are hopping around like that. Sometimes, as you know, you'll be cruising along, and then the standards all go to pieces for reasons that no one seems to be able to explain.
Permalink to CommentThough I'm a chemist, I'm no fan of the "blame the assay" method. However, I do advocate the "ditch the assay" method, if it becomes necessary.
6. tim mayer on August 2, 2005 12:13 PM writes...
"5. For bonus points, how do you tell which numbers you've been getting are closer to the truth - the ones that say your compounds are really active, or the ones that say that they stink?"
Which is why I never do less than four runs on any given experiment. I don't care how much Jmp or any of the DOE programs work, if garbage goes in, garbage comes out.
Permalink to Comment7. jeet on August 2, 2005 1:00 PM writes...
I hate saying "it depends," but what the hell, it depends.
Sometimes you are working with a whole host of assays (in vivo mainly) that are relatively unproven and on a biological model that is mostly theory. Hopefully this means that you are gunning for a drug that will be significantly disease modifying and in a disease that has few good treatment options. Not surprising, but many of the animal and cell assays probably have little to no correlation to the human disease being targeted, leading to failed clinical trials. In this case you are testing both a compound and the underlying theory. I would say this gives you more room to say yes to moving things forward, because the opportunity is more significant and trying to be highly significant in an highly uncertain environment isn't going to work. Plus, first to market is a huge advantage.
Permalink to CommentIf you are developing a 2nd or 3rd generation compound with known human biological response, I would say you need to be a lot more discriminating. The drug is going to have a lot more development (external and internal) and market competition, and you should be aiming for a higher probability of success to compensate.
8. Thomas E. McEntee on August 3, 2005 7:33 AM writes...
Re #2, back in the 1970s, when I was a young PhD process development chemist charged with taking the lab processes for pharamceuticals and pharma intermedidates and getting these ready for 2000-4000 gallon Pfaudler reactors, we viewed any statement by R&D types that they had a "good, large-scale route" as humorous. In retrospect, it wasn't their job to develop efficient large-scale processes. We did, over time, develop decent working relationships with the R&D people that included having them work with us in our environment and us working with them in theirs. This was beneficial in the long run.
Even within our process development environment, we had chemists who were really good with some types of reactions and not others. The good chemists learned to accept their strengths and weaknesses and to reach out to expertise when needed.
Permalink to Comment