You’ll have noticed that scientific discovery often follows a template set by mathematics. A particular result gets generalized to a class, then other fields grow up around the relationship of that class to other classes and around the various ways to make those sorts of generalizations: higher levels of abstraction. Here are examples of something; here are the rules that explain them. And here are the rules about making rules, and how to use the special screwdriver that builds other screwdrivers.
Things tend to get more powerful as they get more general, and we’re seeing that at work in this new paper on automated synthesis from the Jamison group at MIT. A first step in that field was demonstrating that you can use mechanical devices to set up a chemical reaction, and this has been extended over the years to setting up arrays of reactions, arrays under different conditions, reactions that are then analyzed by other coupled machines, and most recently systems that set up arrays of reactions to optimize some particular transformation, analyze the results, and then set up a further round of experiments based the parameters and yields seen in the first set. And this latest work is an attempt to generalize that even more; to produce an automated platform that can handle a whole range of common reactions in this way. That is, to come closer to a Universal Synthetic Chemistry Machine. It’s not easy:
. . .we aspired to create a compact, fully integrated, easily reconfigurable, benchtop system that enables automated optimization of a wide range of chemical transformations. During the initial investigations of such a system, we recognized that several challenges would need to be addressed. These include the chemical compatibility of components and pumping mechanisms; development of a unified, modular system for truly plug-and-play operation; appropriate software for system control and real-time monitoring (using established analytical methods) for automated feedback optimization; and ultimately, integration into a single, small-footprint platform that requires little user expertise with flow chemistry.
The system they’ve built has three modes: automated reaction optimization, automated expansion of a given reaction to a wider range of starting materials, and scale-up of a given optimized reaction. One of the biggest challenges was getting things to work in a plug-and-play manner (and I can well believe it). They’ve ended up developing a sort of universal connector bay, and five different modules (so far) that plug into it: a heated reactor (up to 120°C), a cooled reactor (to –20°C), an LED photochemistry reactor, a solid-supported-reagent packed-bed reactor (, a membrane-based liquid-liquid separator (for extraction). There’s also a bypass connector for unused bays, etc.
The reactions optimized in the paper are the Pall-Knorr pyrrole synthesis (amines with 1,4-diketones), a Buchwald-Hartwig coupling system, a Horner-Wadsworth-Emmons olefination, a reductive amination, a Suzuki-Miyaura coupling, a nucleophilic aromatic substitution, photoredox iminium generation/nucleophile trapping, and ketene generation from an acid chloride followed by cycloaddition with an alkene. That’s a pretty good range of standard reactions. In each case, several variable were investigated (typically 3 or 4), and the machine ran in the range of 30 to 60 reactions on a time scale of hours. The off-the-shelf SNOBFIT module for MATLAB was used for the optimization process itself. Compounds could be produced in up to gram quantities once the reactions were optimized on 10mg scale – as with flow systems in general, scaleup can mean just running the system longer. Here’s how the group sums up (emphasis added):
This reconfigurable system has changed the way we approach experimentation and optimization in several ways. It accelerates the synthesis of lab-scale quantities of molecules and allows investigators to direct more of their efforts toward the creative aspects of research. The system’s generality and ease of use obviates the need for expertise in flow chemistry to realize its benefits. The system also provides a means to optimize and evaluate the scope of a reaction in a matter of hours or days and do so under identical reaction conditions for each substrate of interest, if desired. Transfer of experimental results is now direct, electronic, and seamless; the time-consuming exercise of reoptimizing literature procedures should thus diminish in its frequency.
I highlighted that part because I also emphasize this when I talk about AI and automation. Think about instrumentation such as NMR, LC/MS, and mass-based purification systems. At the beginning, all of these technologies were machines run by dedicated operators (or whole staffs of them), but they have, over the years, become robust enough to where people can use them on a walk-up basis. You certainly need people keeping them running well, and that will most certainly be the case for synthesis machines, but I think that we can now see the way towards such machines becoming reality.
This work shows the optimization of a range of reactions by a single versatile machine. Eventually such systems will be hooked up to retrosynthesis software (which is likewise improving all the time), and pretty soon you’re heading towards that vision I linked to above. Not this afternoon, not next week, not next year. But it’s coming. The need for small-molecule synthesis is large, the time and money spent on it is large, and automation can squeeze a lot of time and money out of the whole process. Our job, as human organic chemists, is to end up on the outside of the zone being squeezed. Prepare accordingly.