Teaching the computer to autonomously plan multistep chemical synthesis has, for over five decades, been one of the biggest challenges in modern-era organic chemistry. Recently, we have disclosed the algorithmic basis of the <i>Chematica</i> retrosynthetic platform and subsequently validated its predictions in improved syntheses of high-value, medicinally-relevant targets. With such examples in hand, we have been expanding the scope of the software to address other needs – in particular, design of routes that avoid patented approaches. At first sight, this can be achieved by simply eliminating specific reactions or intermediates. In practice, however, the machine can then use structurally similar substrates/intermediates or equivalent reaction protocols while still applying the same retrosynthetic disconnections. We have overcome this problem by developing algorithms that (i) track specific bonds along entire computer-generated routes and (ii) allow the program’s user to specify which bonds are to be preserved. Our algorithm is based on the propagation of bond sets from one retrosynthetic generation to another, which is superior to the preservation of entire structural motifs because such motifs may not be unique along more complicated syntheses. Importantly, when bonds corresponding to key disconnections in patented approaches are marked as not-to-be-cut, <i>Chematica</i> is forced to seek alternative and qualitatively different solutions. In my presentation, I will show how this protocol allows the machine to swiftly navigate around patented routes leading to drugs, clinical candidates or natural products, in all cases creating within minutes elegant and chemically viable synthetic solutions. I will also put these examples in a wider context, arguing that the machine’s ability to inspect synthetic options almost exhaustively and to find loopholes in the existing IP might auger a fundamental change in how chemical-synthetic IP is protected and/or challenged.

Navigating around patented routes with the help of computer-driven retrosynthetic analysis

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.