On being a computational chemist in industry

In a recent post on Chemjobber, Lisa Balbes interviewed a computational chemist in the pharmaceutical industry about his job description and the skills that are needed to work as a modeler in industry. And as a computational chemist working on applied problems for almost a decade now (goodness gracious), this gives me the perfect reason to hold forth a little on this topic. I may do a series of posts later, but for now here's what I think is the low down.

Let's get the most important thing out of the way first. It is absolutely important for a modeler to speak the language of the medicinal chemist and biologist. Personally, in spite of being a computational chemist, I always consider myself first and foremost an organic chemist (and I did go to graduate school in organic chemistry before specializing in modeling), using modeling only as a set of tools to shed light on interesting chemical problems. In fact I find myself spending as much time studying the literature on synthesis, physical chemistry, biological assays and protein structure as on modeling.

Computational chemistry is certainly a bonafide field of chemistry in itself now, but especially in industry it's primarily the means to an end. It doesn't matter how well versed you are with a particular technique like molecular dynamics or quantum chemistry, what matters the most is how well you understand the strengths and limitations of these methodologies. Understanding the limitations is as important since only this can help you decide in the end how much you can trust your results - a prerequisite for any scientist. What is key is your knowledge of the chemical system under consideration that will allow you to best choose a judicious combination of relevant techniques. And even this is not as important as the final goal: being able to interpret the results in the language of chemistry that everyone understands, telling your colleagues what it means and how they should now proceed, with all the appropriate caveats and optimism that apply. Understanding and conveying the uncertainty in your methods is as important as anything else since your colleagues need to hear an informed viewpoint that tells them what they are in for rather than a blind prediction.

Unfortunately I have met my share of modelers who think that their expertise in programming or in the intimate working details of one particular method automatically qualifies them to shed light on the details of an interesting medicinal system. Broadly speaking, modelers can be categorized between method developers and application scientists. There is of course considerable overlap between the two and both are valuable but let's make no mistake; in industry the ones who can directly contribute to a project the most are the latter, using tools developed by the former. No amount of training in C++ or in the mathematical wizardry behind a quantum chemical method can prepare you for intuiting the subtle interplay between electrostatic, steric, polar and nonpolar interactions that cause a ligand to bind to a protein with high affinity and selectivity. Much of this comes from experience of course, but it also develops from being able to constantly appreciate the basic chemical features of a system rather than getting hung up on the details of the method.

As we have seen in other posts, a lot of chemical problem solving depends on intuition, an almost tactile feel for how atoms and molecules interact with each other. This falls squarely within the purview of basic chemistry, most of the kind that we learnt in college and graduate school. An ideal computational chemist in industry should first and foremost be a chemist; the "computational" part of the title describes the means to the end. There is no substitute for basic familiarity with the principles of conformational analysis, acid-base equilibria, physical organic chemistry, protein structure, thermodynamics and stereochemistry. Nobody can be good computational chemists if they are not good chemists to begin with.

Apart from these skills, modelers can also bring some more under-appreciated skills to the table. Those who look at protein and ligand structures on the screen all day long usually have a much better sense of molecular sizes and volumes compared to bench chemists. A medicinal chemist might look at a protein cavity and conclude that it's big enough to fit a cyclohexyl group, but a modeler might display the cavity in space-filling interactions and doom any such idea to the realm of steric hell. Unfortunately the kind of line drawings that chemists are accustomed to give a false impression of size and shape, and sometimes simply looking at structures in space-filling mode on a screen can do wonders for deciding whether a particular group will fit into a particular part of a protein. This also makes modelers responsible for something that may need awesome powers of persuasion; convincing your experimental colleagues to regularly come to your desk and look at some pretty pictures (as an aside, modelers may have to play especially nice with their colleagues). Looking at protein structures and molecules all the time should ideally also make a modeler something of an informal expert in structural biology and physical chemistry. Thermodynamics especially is one area where modelers might know more than their organic colleagues because of their focus on the free energy of binding, and I have occasionally productively contributed to discussions about enthalpy, entropy and isothermal titration calorimetry (ITC). In addition, doing structure-based design is always a good opportunity to learn about x-ray crystallography and NMR spectroscopy. You may increasingly find that your colleagues come to you for advice on many structural aspects of their disciplines.

Ultimately a modeler's value to an organization is going to be judged on the basis of her abilities to offer practical suggestions to her colleagues in the language of their own disciplines (as well as the shared language of basic chemistry). The more organic chemistry and biology she knows, the more she will be cherished. The more she empathizes with the particular intricacies of her colleagues' disciplines, the more she will be regarded as an asset. As just an illustration, let me recount a personal anecdote.

I was collaborating with some chemists on a kinase inhibitor project. At one point I thought of a modification to our compound that looked very promising. At the next meeting, here's what I said to my medicinal chemistry colleague: "Jim, there are two modifications that I thought might improve the potency of our hits. One looks very promising, but I have studied your synthetic scheme and I think this modification might be a little intractable, especially considering the cost of your building blocks. On the other hand, here's this other modification which would be my second-best choice, but which you can probably easily install using a Buchwald-Hartwig coupling reaction."

Both me and my colleague were whistling all day long.

Image source