The danger of additivity

When one embarks on the treacherous journey in organic chemistry research, proper planning of experiments is of utmost important. Nothing is more dreadful than going from one experiment to the next while changing two different parameters at a time. If you are a chemistry professor mentoring graduate students, I am sure that you wholeheartedly agree with me and may recall cases when this happened to beginning students in your lab (for example, an attempt to concurrently change a concentration and temperature in a reaction). The core of our work is to ensure that we take a rational approach to incremental learning, which is based on looking for cause/effect correlations while focusing on one variable at a time.

The idea of incremental changes goes beyond running experiments and affects our reasoning by implying that additivity should be the guiding light in reaching sound conclusions. I will provide evidence where being too dogmatic about additivity is counterproductive. As you can see, Klebe and co-workers make an excellent point: if you modify the inhibitor on the top left with a methyl group, you will get a molecule, whose binding affinity to thermolysin is improved only marginally (2.2 kcal / mol gain). If you then modify the same starting point with a carboxylic acid, again there is nothing remarkable (1 kcal / mol gain). But if you now do both of these changes (methyl and acid) at the same time, the result is profoundly better than the starting point (6.7 kcal / mol gain). While the underlying reasons for this sort of behavior are complex, this set of examples speaks to the non-additivity of functional groups and suggests that it is wrong to think about molecules as Lego-like agglomerates of functional groups. Every molecule is in its own class and simple functional group additivity is not always a sound guiding principle. You might then ask: does this imply that the vast majority of medicinal chemistry research is misguided? I don’t know. Maybe it is.


7 thoughts on “The danger of additivity

  1. Recently I came across a couple of papers (like this and blog articles (like this about Design of Experiment (acompletly new subject to me) in which the authors discuss the fundamental flaws of One-Variable-At-Time approach.
    It seems hardly applicable to a fragment based modification work, but very powerfull in the development of reaction conditions (so maybe my comment is not relevant about this post).
    Anyway, as a chemistry student with any knowledge in these subjects, I am interested in your opinion about OVAT approach vs multivariate approach.

    • Thanks for bringing up this point. Of course, there are excellent chemoinformatic algorithms such as the ones you refer to (and I am glad you mentioned these resources). However, without technical knowledge of how to do this properly, one has to rely on more conventional approaches. But I do think the following question deserves through analysis: what can we do to seamlessly incorporate modern chemoinformatic approaches into experimental design on a day to day basis?

      Having said all of this, I think that the fragment lesson that is presented is particularly instructive and highlights what is “emergent property” when unrelated parameters are considered at once.

  2. Nice post! It reminds me (well, almost everything does) to the Atoms in Molecules approach by Bader. Where does an atom end and a functional group begins? There is a somewhat arbitrary definition of what a full FG is as evidenced above, i.e. that whole group in red is a full new group as opposed to just the ‘sum’ of the methyl and the carboxylic acid but it can only be evidenced when the three are compared.
    It is a great idea to keep this in mind but do you think it can be envisioned as an a priori tool or just after all combinations have been considered? (yikes, I can’t write in English today! sorry)

      • Sort of like CoMFA? Because it does seem to me like the field of the target is key in these cases although I understand that displacement of water is the driving force in this particular case (I haven’t read the full article)

  3. The nonadditivity may reflect differences in conformational preferences when CH2X or CH(X)Y are linked to the amide nitrogen. I’ve used linked a blog post that may be of interest as the URL for this comment.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s