Citation:
Tyler, J.L.; Trauner, D.; Glorius, F. Reaction development: a student's checklist. Chem. Soc. Rev. ASAP. https://pubs.rsc.org/en/content/articlelanding/2025/cs/d4cs01046a
Summary Figure:
Background:
This is a bit of a weird one today. I was recently shown this paper from the minds of Trauner and Glorius, which they describe as "a helpful guide for synthesis development, allowing you to thoroughly investigate the chemistry in question while ensuring that no key aspect of the project is overlooked." Now, Trauner and Glorius have developed a lot more reactions than me, but this checklist confuses me, because it skips a lot of the steps and seems very out of order. I think it's worth looking at (a) what points they make and (b) how to actually go about doing what they prescribe.
How it works:
The paper is broken down into 8 sections, which I have copied below:
0. How to discover a reaction
(a) develop it from the older literature.
(b) accidentally
(c) by analogy
(d) by necessity
(e) by machine learning
1. Take stock
(a) Have you made the desired compound?
(b) are all reaction conditions necessary?
(c) Is the reaction reproducible and reliable?
(d) Have you found the minimal synthon and retron?
2. Kinetics and Thermodynamics
(a) Can the rate law be determined?
(b) Can you increase the rate?
(c) Have you established the steric sensitivity?
(d) Have you identified the reaction driving force?
(e) Can you bias an equilibrium?
(f) Is the intended product stable?
3. Mechanism
(a) Does the proposed mechanism require evidence?
(b) Can the intermediates be identified?
(c) Can the reaction pathway be computed?
(d) Is there an alternative route to the active species?
4. Optimization
(a) Has the yield been fully maximized?
(b) Can reaction selectivity be achieved?
(c) Can the protocol be more efficient?
5. Catalysis
(a) Can the reaction be catalysed?
(b) Have you identified the catalytically active species?
(c) Can you catalyse the reaction enantioselectively?
(d) Is the catalyst turnover optimal?
6. Scope
(a) Is the substrate scope diverse?
(b) Have you demonstrated functional group tolerance?
(c) Have you defined the parameters of reactivity?
7. Applications
(a) Can it be used to access target compounds?
(b) Can it be used for late-stage transformations?
(c) Can the products be derivatised?
(d) Can it be used for bioconjugation?
(e) Can it be used in materials chemistry?
I do wish they had made an actual checklist, but here you go. I'm now going to go through each of these and discuss what I think makes sense and what doesn't. This is going to be long.
0. How to discover a reaction
(a) develop it from the older literature.
(b) accidentally
(c) by analogy
(d) by necessity
(e) by machine learning
To develop a reaction from older literature, you go through old papers that maybe had poor conditions (meaning expensive/difficult to synthesize/hazardous reagents) or poor yields/scope and combine the general idea with more modern understanding of the chemistry. For example, Ullmann coupling to dimerize aryl bromides has been around for over a hundred years, but the discovery of organometallic chemistry enabled much better catalyst design, leading to selectivity that made the reaction much more valuable.
Discovering reactions accidentally does happen, but the important thing is you always have to be expecting it. Discovering something accidentally is hard because if your reaction fails, you have to characterize the byproducts to discover what happened. 95% of the time, you discover that either the byproduct is something obvious and not interesting, or the data is so messy that it isn't readable. In the example the authors give of an electrocyclization that gave an alternate product, the researchers had to push through the failure of the reaction to characterize the synthesized product, even though it doesn't obviously push them towards their original goal.
Discovering a reaction by analogy is probably the most common method of developing a reaction. I personally think of this in terms of elementary steps- take known elementary steps and combine them in new ways or with different reagents to generate new products. For example, once the oxidative addition-transmetalation-reductive elimination cycle was developed for cross-couplings, swapping out different metals to transmetalate with gave new reactions. Then, swapping out transmetalation for ligand association gave Buchwald-Hartwig coupling reactions.
I hate the idea of inventing reactions "out of necessity" because (a) this fails all the time and (b) it sounds like a threat "you're going to make this reaction...or else". We need to develop a reaction to oxidize methane to methanol, it's not like the chemists working on that problem haven't tried hard enough. Either (a) you need to employ one of the other methods (usually either from older literature or by analogy) or (b) you just screen conditions until you succeed, which is how you end up with Phil Baran screening literally a thousand reactions until it works.
Machine learning is also not a method for discovering reactions, it is a tool. Either it discovers something accidentally, or you are screening conditions intentionally and are therefore discovering by analogy. Honestly, both this and the previous method rely a lot on brute force, which is a valid option, but probably not your first choice.
On to step 2 (which they call step 1)
1. Take stock
(a) Have you made the desired compound?
(b) are all reaction conditions necessary?
(c) Is the reaction reproducible and reliable?
(d) Have you found the minimal synthon and retron?
Yes, it is absolutely necessary to make sure that you have made the desired compound. This normally is obvious, but there can be some addendums. For example, if you expect your reaction is enantioselective, you should check that it is actually enantioselective early on in the process. If your reaction could make multiple regioisomers, you should check which one is formed early on. This will impact the type of products you make and if the project is viable. I dislike how the authors suggest X-ray crystallography as the first technique, as cheaper and simpler techniques like NMR and HRMS are the first line of defense, and more costly but thorough techniques should only be employed if necessary. For example, IR should only be needed if there is a question of a certain functional group being formed, or polarimetry to verify enantioselectivity (although chiral HPLC/SFC is much better).
Similarly, the reaction needs to be reproducible and reliable. Frank Glorius an author, so obviously he pitches the use of his sensitivity screen, but I think one of the most basic things that needs to be done is to run the reaction a second time, exactly the same, and verify that the results are the same (and ideally with a second pair of hands).
Where I think this list starts to get weird is how the optimization table (usually Figure 1 in most papers) is broken up into multiple sections. These questions:
Are all the reaction conditions necessary?
Is the reaction reproducible and reliable?
Has the yield been fully maximized?
Can the protocol be more efficient?
Can reaction selectivity be achieved?
All are answered by the same thing, which is an optimization table. Finding the set of reagents that maximize product formation and minimize byproduct formation will answer all of these. Note that I'm not saying these questions are the exact same as each other. For example, to make the protocol more efficient part of optimization is finding the minimum amount of reagents necessary for a transformation; if 2.0 equivalents and 4.0 equivalents of a reagent both lead to 78% yield, use 2.0 equivalents. A different part of optimization is finding if the yield can be increased; if 2.0 equivalents leads to 45% yield and 4.0 equivalents leads to 78% yield, use 4.0 equivalents. The experiment necessary to answer both of these questions is to run the reaction with 2.0 equivalents and again with 4.0 equivalents and measure the yield.
I don't really understand what they mean by "find the minimal synthon and retron," their description sounds like reaction scope based on the line, "indicate whether simpler starting materials or reagents can be employed under the same conditions to broaden its applicability"
I'm now at over a thousand words and nowhere near the end, so I'm going to break this up into 3 parts:
1. Finding a reaction and optimizing
2. Identifying the mechanism
3. Scope and applications
Key takeaways:
There are a number of different ways to discover a new reaction, and none of them are easy. They either require (a) Good chemical background and a thorough and/or lucky lit search (b) hard work in the lab, perseverance, and a little luck (c) A lot of work in the lab, perseverance, and a little luck. Once you do discover a reaction, optimization answers a lot of individual questions about how to get the reaction to work, which is why it's necessary.
(Part 2: https://alwaysbecoupling.blogspot.com/2025/02/reaction-development-checklist-part-2.html)
(Part 3: https://alwaysbecoupling.blogspot.com/2025/02/reaction-development-checklist-part-3.html)
Comments
Post a Comment