Machine Proposed Synthesis of Tamiflu
A 6 step machine proposed synthesis is outlined in the handrawn figure attached.
The software suggested all disconnections, the SMs, and the reactions to be used. More detailed information about this synthesis, the individual reactions, and the project in general can be obtained by contacting us at the email address provided below.
About this synthesis:
STATS:
The longest linear sequence is 3 steps.
There will be a total of 6 Steps if the SM(s) are not commercially available.
Compound A can most likely be made in 1 step depending on what is commercially available. Compound B can most likely be made in 2 steps depending on what is commercially available.
- The first proposed reaction is a Grubbs CM followed by an RCM with the second gen catalyst.
*Note, CM reaction selectivity is difficult on some substrates, once the CM reaction goes, most RCM reactions are very selective, fast, and can be easily accomplished. For more info on why this reaction was selected by the software, please contact us directly at the email address provided below.
- The second step is the aziridine formation (J. Am. Chem. SOC.,Vol. 116, No.7, 1994)
There may be issues with selectivity here, but we predict that the favored reaction will be that of the less hindered double bond, which is where we want the aziridine to form.
- The third step is the asymmetric opening with TMSN3 (http://pubs.acs.org/doi/abs/10.1021/ja061696k)
Potential Issues:
- This synthesis has not been tried in the lab.
- Regioselectivity for the aziridine formation is expected, but as this hasn't been tried in a lab, it cannot be guaranteed.
To clear up any potential confusion...
- the custom conditions for the novel tandem CM/RCM reaction have not been made available online, but more info can be obtained by contacting us directly. With the help of a chiral-co catalyst and specfic reaction conditions, the software predicts the tandem reaction will be successful. In any event, due to the highly adaptable nature of the synthesis, should it fail, Tamiflu can stil be reached via the alternative methods utiliing the same later reaction steps, and in 5 steps.
--------------------------------------------
It's important to note that:
The software proposed multiple alternative syntheses with more promising predictions of success. The alternative syntheses were slightly longer, but still shorter than those currently published. However, this synthesis was the most interesting for our purposes, thus we've chosen to make it available on figshare.
Positives:
The synthesis is highly convergent, far shorter than those currently published, utilizes new disconnections, and provides highly modifiable/adaptable intermediates for customization.
Summary:
This synthesis was produced by the reaction prediction software project at Intell.Tech.Scientific.
www.IntelliTechScientific.com
The reaction prediction software project utilizes machine learning, molecular physics, and synthetic organic chemistry to predict the outcome of novel reactions, the outcome of known reactions on novel substrates, assist with synthetic planning, and other reaction planning related tasks.
This synthesis was significant to us because it showed an improvement on machine selected SM, and overall improvement on reaction selection for a total synthesis (though there were still some issues), and continued adherence to highly convergent and adaptable syntheses, which are what we are trying to achieve.
At this time we are looking for negative data from synthetic organic chemists to help train our custom built neural networks.
For more information, please send an email to Jamie@ITSMolecularDesign.com.
update: http://figshare.com/articles/Tamiflu_Synthesis_Part_2/904942
we've also just published a slightly altered version of the Tamiflu Synthesis, which provides a more traditional synthesis. It is still 6 steps overall, but now has 5 in the longst linear sequence, and would be useful if the tandem Grubbs CM/RCM doesn't work, or if the CM/RCM reaction produces a mixture of products due to incorrect predictions regarding the coordination and chelation/regioselectivity for the CM, and/or incorrect predictions regarding the ease with which a chiral resolution could be employed at the end of the third step (in the originally proposed synthesis) to separate any mixtures of isomers, should the predictive models prove incorrect in regards to the coordination/chelation/regioselectivity, which it's believed may lead to the formation of a majority of only the desired product.
In the altered version of the synthesis, the stereochemistry can be set after the ring has been formed via the exact same methods proposed in the original synthesis. However, if the tandem Grubbs reactions proved successful but produced a mixture of products as a result of issues witht the coordination/chelation models, then a chiral resolution would be an ideal way to retrieve the desired product.
