Bringing Computer Assisted Organic Synthesis to Medicinal Chemistry

By David Clark, Stephen Penrose, Charles Baker-Glenn & James Mui.

This article was adapted from a post that appeared on Eureka, Charles River’s science blog.

These days, it’s almost unthinkable to undertake a car journey of any complexity or distance without employing a GPS (or Sat Nav) to help us take the most efficient route to our destination. In a similar way, when a chemist working in the lab embarks upon the synthesis of a compound, he or she follows a “synthetic route”— a series of steps by which a chemical compound is assembled from smaller, less complex bulk chemicals. These routes can be devised based on the chemist’s experience and by searching the scientific literature to learn from the experience of other chemists who may have attempted similar synthetic paths in the past.

But what if there was a “GPS for chemists”? A piece of software that could look at their destination (the drug) and then propose one or more routes to it—perhaps the most efficient or cost-effective? Such a tool could help to speed up compound synthesis by mining a much greater wealth of stored chemical knowledge and avoid “blind spots” or biases in a chemist’s knowledge. It’s a tough and pot-holed road to bring a drug to market, with costs reaching $2.6 billion according to a recent study, so any help along the way would be welcome!

Standby to enter the world of CAOS – Computer-Assisted Organic Synthesis. In fact, the idea of using computer software for this task dates back to the pioneering work by Corey and Wipke in the late 1960s. But over the last few years, the methodologies have been advancing and several commercial products are now on or approaching the market. These promise to put CAOS in the hands of the lab-based chemist.

CAOS software packages can be used in prospective route planning with or without expert scientific input. The packages rely on de-novo techniques, using rules set by computational chemists to predict and rank the most suitable retrosynthetic transformations to apply to a drug molecule. The scope of reactions in which an expert chemist will have experience will be narrow compared to the breadth that machine learning can apply. Certain commercial CAOS software packages have been highlighted in a review by scientists from AstraZeneca.

One of them, ICSYNTH ,is a piece of software that creates rules based on the findings of many years of chemistry research. These rules tell the user what is and is not possible. In journey terms, this can be likened to an understanding of which roads are clear and which ones lead to traffic jams. The user can choose a preferred route based on their needs. For example, one user might take a route that is cheaper (choosing the more fuel-efficient vehicle), one that is faster (choosing the shortest route) or one that is less speculative (choosing a route that is used more frequently). A report in the literature describes the evaluation of ICSYNTH by comparing the program’s performance in predicting new ideas for route design for the synthesis of some compounds of interest. The program’s suggestions were
compared to the output of historical brainstorm results from project chemists, as well as literature data. The overall conclusion of the exercise was that ICSYNTH can add appreciable value to the performance of a team of R&D chemists and indeed it is in routine use at AstraZeneca (e.g., in route design for AZD-4635, an adenosine A2A receptor antagonist).

ICSYNTH has the potential to be a powerful assistant to the laboratory chemist helping him or her to find, and then choose between, multiple routes to a drug target. In addition, such software can also help with prioritisation of target compounds, i.e. deciding whether to make molecule A or molecule B.

By enabling better decision making, a chemistry-aware GPS should reduce the time required for each cycle of making and testing new drug molecules. The ultimate goal at Charles River, and for all drug discovery scientists, is to bring more medicines to market that make people’s lives healthier and happier. Planning the journey to get us there quicker and smarter can only be a positive thing!

More information

DeepMatter Group acquired InfoChem based in Munich, Germany. InfoChem has more than 30 years; experience in the development and integration of sophisticated software tools for the storage and handling of structure and reaction information. The company main activities involve the production of synthesis planning and reaction prediction solutions using ICSYNTH and the automatic extraction of scientific information from text and images. For more information click here www.infochem.de