July 10, 2007
by Ryan Sasaki, NMR Product Manager, ACD/Labs
As I mentioned before in this blog, I have spent the last year and a half visiting with medicinal and process chemists, as well as NMR Spectroscopists who support chemists in an open access environment. I‘ve been trying to understand how NMR software is being used in these environments and how things can be improved in this world.
A few weeks ago, I blogged about how chemists use NMR and I also questioned the information (or lack thereof) in archived data, and spectra in reports.
I am going to riff a bit about what I have learned about compound registration. In the great majority of medchem groups I have spoken with, a compound is synthesized, an NMR and sometimes an LC-MS is run, and then a registration form is filled out. This registration form consists of a check box where the chemist selects which experiments were run and a statement such as, “the structure is consistent with the data collected”
My question is, what happens with this data? Generally this data will be pasted in a chemist’s lab notebook, or maybe even in their ELN if the organization is that far along. But for the most part this is just data. Whether it’s a printout or an electronic file, it’s generally just the data. Maybe there is a structure attached, and some chemists will diligently assign their spectra to their structure on their printout.
But the issue I see here is that there is no consistency. There is no mandate at the top that is telling the chemists to provide support of a proof of structure. A 1H NMR and mass value cannot provide a full proof of structure…but how exactly is a simple statement like, “the structure is consistent with the spectrum” validated? Certainly it’s not validated from a paper printout of a spectrum.
I hope people realize that my intention is not to call out chemists. The chemists are not the problem here. Their job at this particular stage is to make compounds and fill out a registration form (forgive me for being overly simplistic). Their job is not to fully assign their data and prove that at least there is nothing in the data that suggests an inconsistency. Again, a 1H NMR is not going to be a proof of structure. But it was run for a reason, and that reason should be validated.
I’ve talked to many chemists who are very diligent about their handling of NMR data. They will use our software and fully assign their data to a chemical structure and ensure that there is no inconsistency. They do this not because they have to, but for their own piece of mind, and because they believe it is the responsible thing to do. But again, the problem is consistency. For every chemist who is diligent about their NMR data handling, there are several more that don’t do it. In fact, there are perhaps some who run the NMR but don’t look at it. The identity of their compound is confirmed by a mass value, and in the environment they are working that is perfectly acceptable. I personally don’t blame them for not spending the extra couple of minutes to fully assign their data. It’s not their job!
Maybe I am out of place…maybe I am misunderstanding the environment or the reasoning behind not taking this direction. Perhaps this just isn’t the right time to bring this issue up. But in the age of electronic lab notebooks, to me, it seems like a perfect time to discuss and implement this.
But in an industry that is so heavily regulated, it blows my mind that a chemist’s decision to register a compound is based on good faith.
Perhaps an expert in this area can clear things up for me and put me in my place. I am simply trying to espouse the benefits of our software. I believe assigning NMR data electronically to structures has major benefits.
P.S. There’s another benefit. Imagine chemists in an open access environment fully assigning all of their NMR data and then sending it off to an NMR Spectroscopists. This will create a purgatory database for the NMR spectroscopist to evaluate. Once quality checked the spectroscopists can build a database of all this information and in turn, vastly improve the NMR prediction capabilities within the organization.
The answer is very simple. The purpose of any pharmaceutical industry is to bring a drug to the market. For this purpose thousands of compunds are synthetized. What’s important is that the one compund going to the market is carefully certified. Validating all the products is a waste of time and money. It’s also true that acquiring a spectrum and storing it without giving a look to it is a worse waste of time. You are simply worried because you want to sell your software. The more time researchers spend in quality assurance, the less drugs they invent!
Thanks for sharing your thoughts Giuseppe.
I agree that the purpose of pharma companies is to bring drugs to market.
Perhaps validating all products is a waste of time and money…but in doing so companies can build a meaningful knowledge base of their compound library…but even more importantly they can probably reduce the number of incorrect structures in their registration database.
I am not simply bringing up this issue because I want to sell our software. In fact, many, many companies already have our software deployed to all of their chemists in an open access environment and along with it the ability to electronically assign all of their NMR data. Some of these medchem groups are actually doing this, others want to start doing it, it’s just that these groups seem to be in the minority.
From my experience as a chemist in the pharma industry, I learned very early on that there is an incredible amount of time, money, and effort spent on quality assurance. I’ll counter your point by saying the less time researchers spend on quality assurance, the lower quality products they’ll produce.
Ryan, nice topic! One that’s especially interesting to me, in light of my previous career as a spectroscopist supporting medicinal and combinatorial chemists. I’ll keep my comments focused by responding to the first responder, Giuseppe. Two of Giuseppe’s comments resonate quite negatively, at least in my mind:
1) “Validating all products is a waste of time and money” This is only true, if you have a very narrow view of what “validation” is. No, a full fledged proof of structure isn’t needed in discovery. But, confirming that the chemistry you performed worked IS necessary. Why? Because if you synthesize the wrong structure, and you steer the SAR of a project in the wrong direction, you screw the project. I’ve seen this happen, and it’s not fun…
2) “The more time researchers spend in quality assurance, the less drugs they invent!” I simply don’t agree with this statement. Yes, successful discovery organizations foster an environment of creativity. And yes, GxP requirements are of course not appropriate for discovery. But, as a scientist, I can’t accept that no validation is a good thing. Let’s take Giuseppe’s statement further, and suggest that chemists stop describing their syntheses in a lab notebook. Furthermore, let’s suggest that there be no registration process at all…just mix and test! How long would it take for a drug to be discovered in that environment?
My summary assessment of the problem is, like most things in life, a careful balance must be struck between a quality and speed.
Finally, one more comment: as blog readers/responders, let’s dispense with the accusations of hidden agendas. Yes, Ryan and I have “day jobs.” But I am not responding to this BLOG to enhance my potential to sell more software. I’m motivated to respond because I’m interested in the discussion, period. I won’t speak for Ryan, but I’m confident he’ll confirm what his interests and worries are.
Let me explain my position. I am convinced that keeping a lab notebook is fundamental. I am convinced that following a few rules is fundamental, because it allows every researcher to read and understand what’s written in somebody else’s notebook. I am also convinced that acquiring and assigning NMR spectra is a must. The concept of “validation” is something that I can’t accept, however. What happens when the “validator” gives a false positive? What happens after one year of using a software validation? It happens that the chemist has forgotten how to read an NMR spectrum.
The point “a wrong structure can steer the project in the wrong direction” proves nothing. Obviously, only a compound with high pharmacology activity has such a potential, and not all compounds have this high activity. The subject of my first post was the thousands of compounds with no potential, whose only future is the “mythic” corporate database.
Thanks for your response and clarification, Giuseppe. However, I think you may be misunderstanding my point somewhat.
Please note, I am not proposing “automated” validation of NMR data by a software (or taking the data out of the hands of the chemist). Not in this particular post. I am specifically talking about the chemists themselves doing complete assignments in an electronic format. In many labs, that is not necessary and the confirmation of a registered compound has the potential to be only loosely based on the data. Again, some chemists will fully assign their data but others will only look for one peak, or might depend solely on the mass value.
If a mandate is in place at the top, that medicinal chemists HAVE to assign their data electronically and link to that data in their notebooks…isn’t that a step forward in improving the quality of the compounds in the registration database?
Wouldn’t full electronic assignments between a spectrum and a structure, in your words, “allow every researcher to read and understand what is written in somebody else’s notebook” even clearer? If you agree that acquiring and ASSIGNING NMR data is a must…then I think you agree with my overall point…Chemists should be asked to fully assign their NMR data to ensure quality.
Whether or not that is done on paper or electronically depends on preference…but I think in this day and age, having this information in a digital format is the way to go. It’s easier to store, link to, and retrieve.
As a software company we have worked very hard for years to try and make the process of assigning multiplets to atoms in a chemical structure as easy and seamless as possible. The goal is to make it faster than it can be done on paper. I think we are close if not already there.