Dykellic Acid (Machine Synthesis)
This synthesis of Dykellic Acid was produced by the beta version of our reaction prediction software. This synthesis has never been tried in the lab, thus there are some potential problems with it. Still, there are aspects of this synthesis that might be useful to others.
How this was accomplished: We're using a series of neural networks that I've created to power our reaction prediction software. We are still in the very early stages, though this project has been ongoing for over three years.
The software successfully carried out a retrosynthetic analysis, it successsfully proposed novel reactions/conditions, and successfully calculated the perceived risk of each step (though we cannot tell how accurate those numbers are without first trying the proposed reactions in the lab). However, it is important to note that the parameters analyzed (the list of which is proprietary) were quite extensive, and we are happy that we were simply able to get the software to run the calculations and produce what seem to be reasonable numbers at first glance.
The numbers are:
Step 1: Grubbs = 85% probability of success
Step 2: Chiral Reduction = 92.7% probability of success
Step 3: Lactonization 47.6% probability of success
Step 4: Oxidation = 98.75% probabiliy of success
Step 5: Tebbe Olefination = 25% probability of success. The software is confident that the Tebbe Olefination would selectively reduce the alpha-beta unsaturated ketone over the lactone.
There were some problems with the identification of promising starting materials and reactions, and some of the steps had lower probabilities of success than others for obvious reasons that point out weaknesses in our training data. However, for the most part, the software did accurately recognize problem reactions, thus we are very happy with this result, and we will be continuing to work to improve the areas of weakness with this software.
Since the 16th of March, when the above proposed synthesis was originally completed, the software has produced a number of other total syntheses. Currently we are working on increasing the probabiity of success for each step, improving the selection of starting materials, and improving the software's ability to suggest appropriate reactions and reaction conditions.
I did not include the original print-outs from the software, as that information contains proprietary data. However, those who are interested in learning more about this software (specifically those who work in the areas of physical organic chemistry or synthetic organic chemistry, that would like to collaborate), can make an application.
Please note that at this time, we're currently not accepting any additional investors, and collaborators must pass our free certification exam at an approved testing center near them before they will be given access to proprietary infromation or provided with a leadership position on the team. You will also be asked to sign non-disclosures, etc.
Contact:
Collaborations@ITSmoleculardesign.com
You can learn more about this project at:
www.IntelliTechScientific.com
