Merck Synthesis Challenge 2024 Route Report- Top 20!

In some very cool news, my team for the 2024 Merck Synthesis Challenge placed top 20. I think it'll be fun to share the route we came up with, the routes we considered but ultimately rejected, and my thoughts on the process. 

For those unfamiliar with it, Merck kGaA (the European one) has hosted a competition loosely every year or two where teams around the world have 48 hours to propose a synthesis of a target molecule. Teams must write out each step as explicitly as possible, with the eventual top routes being attempted by a CRO with as little modification as possible. This means that if you want to make a C-C bond between an ester and an aldehyde, you can't just draw an arrow and say "aldol"- you have to look up precedent, analyze conditions, and justify your choices.

There is also a great summary of the history of the challenge, the goals, and how they design it. The article is open access and worth a read: https://onlinelibrary.wiley.com/doi/full/10.1002/anie.202317338 

Here are the first four compounds, for reference:

 

I also participated in the 5th compound challenge, where we placed in the top 75 teams (out of ~600).  This was the molecule, and here is our route:

 

 

 

Yields are based on the literature conditions, so probably drop 20% off of each step, but overall our route was well-precedented and I'm still pretty proud of it. 

Obviously, that's old news, so here's this year's molecule:

 

It's a nice moderately difficult challenge. There's a 6-6-5 core with a tertiary amine and an internal alkene. 3 stereocenters total, two of which are contiguous. There's a ton of natural product alkaloids and previous total syntheses of molecules that look like this:

 

However, there aren't any known compounds that I could find that have this type of ring system, saturated or not. This means there is a lot of known chemistry to make these bonds in a general sense, but no ultra-clever cascade reaction that forms the core in a single step.

Retrosynthesis:

Our team pretty much immediately saw the following disconnection:

 

I imagine many other teams did too. That ester tail would be fairly difficult to use in earlier steps, and asymmetric reduction of a ketone is one of the easiest and most reliable ways to set an alcohol stereocenter.

Doing some simple consonant/dissonant analysis points out these two key disconnections: 

If we look at the molecule holistically, there are 3 key challenges:

1. The two contiguous stereocenters

2. The tertiary amine

3. The internal alkene

By far, the biggest challenge is the two contiguous stereocenters. 

After we saw the molecule for the first time, we split up to brainstorm routes solo. Most of this time was spent on trying to find good ways to form key bonds, then checking if the structures were feasible. 

Here are some of our most promising initial ideas that ultimately didn't get pursued:

One fairly clever idea we had was to use an intramolecular Diels-Alder reaction to set multiple stereocenters and form one of the rings. We did some quick modeling (pen and paper style) and we were pretty sure that the stereocenters would be set correctly. However, making that diene didn't look trivial, and finding a way to reduce the alkene and then oxidize the amide to the enone and then reduce the amide seemed like it would be a lot of steps. 

Another route that I was really working on was to abuse the fact that one stereocenter was allylic and do a stereospecific cross-coupling, as setting the stereocenters of C-O bonds is generally easier than setting C-C bonds. However, the route ended up being too many steps (~16)

 

Another option was to do an enyne metathesis. Someone in our group found this precedent:

 

We thought that we could then use those two ester handles, change them into NHP esters, and do a doubly decarboxylative cross-coupling (see Baran). We made multiple passes on this route, and got it to a pretty good place, but ultimately there were too many unknowns. For example, could we cleave the exocyclic alkene to access the ketone? 

We also looked at variants of an asymmetric Sakurai allylation, again taking advantage of the allylic stereocenter. We were pretty sure you could oxidize that alcohol to the ketone, then do an asymmetric reductive amination. However, the precedents for the allylation weren't perfect, and the reductive amination wasn't clear either.

After about 18 hours of brainstorming (I'll note that we used a discord server and a miro board), we also had a couple of other key components.

Key Component #1: The best way to make the internal alkene was probably via Grubbs metathesis. 

A lot of our routes relied on reactions that looked like this:

or 

or

 

And there's a few key reasons why. First, Grubbs is a really great method. There aren't a ton of reactions able to couple two fragments together and form an alkene. We wanted to avoid Wittig and similar olefination reactions because we already have a ketone and it might lead to side products; there also might be other sensitive functional groups. If we look back at the consonant/dissonant analysis, this is one of the two places in the molecule where the disconnection clashes. Second, allylation of an amine is a simple reaction that lets us protect the amine for some steps but then use that "protecting group" as a useful building block. Third, this let us focus our energy on the two contiguous stereocenters by solving the other two challenges in one step. 

The other possibility we looked at was an aldol condensation, like this:

 

Which would probably also work, but we considered it a strict downgrade over the Grubbs routes. This aldol would probably be accomplished by taking the allyl amine and doing oxidative cleavage to get to the aldehyde, which seemed more concerning than the route we did end up taking. 

Key component #2: Proline is a great building block. 

This is proline:

 

This is the target intermediate, with the atoms in proline highlighted in red:

Proline contains one of the stereocenters we need, plus a one carbon unit for functionalization with the other sterocenter we needed. There are tons of commercially available proline building blocks with different functional groups.

3. The A ring was easy to acquire

 

Because the dimethyl quaternary carbon was right next to a ketone, it made it really easy to find either building blocks or direct synthesis routes. If possible, we did not want to have to construct the A ring piecemeal, we wanted to add it on in one step. We didn't know if we would want the saturated ring or the enone, but both were accessible.

 However, this leads to the central problem with this molecule. The A ring has 6 core carbons, and proline has 5 core carbons + nitrogen. Either you need to delete the carbon on the carboxylic acid of proline, or you have to construct the A ring from an acyclic building block. 

Then, one member of the team (not me) proposed this very clever idea:

 

This is a Robinson annulation between methyl isopropenyl ketone (a derivative of MVK) and acetylated derivative of proline. This step doesn't set the second stereocenter, but it does form the C-C bond using two very accessible building blocks and no likely side reactions. We considered it likely that we could do an asymmetric hydrogenation of the enone to set the other stereocenter, and then a little more elaboration would finish the molecule. However, this was assuming the Robinson annulation worked. 

At this point, you may go, "Hey, that's a C(sp2)-C(sp3) bond. I bet there's a cross-coupling reaction that would make that"

 The tricky part about this is that the obvious reaction is to use the carboxylic acid of proline as a radical source, plus a vinyl halide, and do some sort of cross-electrophile coupling reaction (see Baran, MacMillan). 

The problem, however, is that you racemize the stereocenter when you decarboxylate:

 

We had already discussed this repeatedly, and unfortunately there is no current precedent for an asymmetric reaction that sets that stereocenter in a stereoconvergent manner either. I imagine there are groups working on this, but there is no precedent right now. 

However, there is more to chemistry than cross-coupling (Blasphemy!) 

I searched up some variants of this intermediate in reaxys:

 

And this:

 

Has been made before!

We found this interesting precedent that uses sparteine as a chiral ligand for copper-catalyzed conjugate additions with alkyllithium reagents. 

 

Link to the paper: https://pubs.acs.org/doi/10.1021/jo035845i

Citation:  Dieter, R.K.; Oba, G.; Chandupatla, K.R.; Topping, C.M.; Lu, K.; Watson, R.T. Reactions and Enantioselectivity in the Reactions of Scalemic Stereogenic a-(N-Carbamoyl)alkylcuprates. J. Org. Chem. 2004, 69, 3076-3086. 

This was huge for us, because now we had a clear path forward. Another member of our group with experience in asymmetric hydrogenation found the perfect reduction:

 

And then we had our two contiguous stereocenters. 

 At this point, I'm going to show our full retro synthesis:

 

Now walking through the forward direction step by step:

 

We found this nice precedent to make the exact vinyl iodide we needed from commercially available material. 

 

Then, we do the asymmetric conjugate addition we found. The only difference between our step and the published step is the two methyls alpha to the ketone, which in my opinion will only improve the reaction by removing an enolizable proton. 

 

Then, we do a copper-catalyzed asymmetric hydrogenation, which in the precedent had been performed on similar beta-substituted cyclic ketones. 

 

 

These next two steps go together, because it is a classic Eschenmoser's methenylation. We had this in our back pocket as the simplest way to make the exocyclic alkene before a Grubbs metathesis. 

 

 

Then, we have to swap out the Boc protecting group for allyl. These conditions are both extremely common. 

Then, we do the much awaited Grubbs metathesis to get to the key intermediate!

We then do a CBS reduction to set the alcohol stereocenter. There are multiple ways to achieve this, and honestly conformational selectivity would likely enable this with just sodium borohydride. However, there is a bulky quaternary center on one side of the ketone and an sp2 carbon on the other, so this is really set up perfectly for the CBS. 

 Finally, a classic esterification with the acyl chloride. You can use the chloride, the anhydride, or Stieglich conditions with the acid. This is a generally simple transformation. 

Our synthesis has 10 steps, all linear (LLS = 10), and an expected overall yield of 21%. 

Strengths: This route is highly feasible. Every step is well precedented, and frankly only the key copper catalyzed asymmetric conjugate addition seems sketchy. This route should be enantioselective and diastereoselective. 

Weaknesses: Unfortunately, this route is completely linear, so it may have a higher number of steps than a more convergent route. Having the conjugate addition as the second step also serves as a bottle neck- that step is probably not scalable, so we had to adjust all of our quantities for small scale. That step is probably not "easy" either, it's no dump-and-stir.  We also didn't really pay attention to how "green" our chemistry is- there's a lot of DCM in here. On the plus side, we did use copper instead of more expensive transition metals (except for Grubbs).

I'm very proud of top 20. Huge shoutout to my team, "If you like pinacol couplings", as this is all of our work. 

Based on trends, last year we were top 75, this year top 20. Next year we'll be top -35!

 

If you liked this, please share it around or leave a comment. I'd love to hear what people thought of the route or what other routes people submitted.