Introduction to ICSYNTH

ICSYNTH and ICFRP: new pathways to innovation

Finding a route to a target molecule has traditionally meant searching databases of known reactions: a method that by its very nature will only find science that has already been published. More recently, computational approaches have allowed chemists to search for novel reaction pathways using either algorithmic (rule-based) or artificial intelligence (machine learning) approaches. However, while these are undoubtedly a step forward from simple database searches, both methods have their drawbacks. A rule-based algorithm is only as good as the rules it follows, and generating those rules manually is a time-consuming process. Machine-learning can offer more novel insight but is often an opaque process: the artificial intelligence is a “black box” that does not let the end user see the rationale behind its decisions. What if you could combine those two approaches, using the power of machine learning to generate understandable chemical rules automatically? And if, rather than relying on published chemistry, you could generate new algorithms from your own proprietary reaction data?

ICSYNTH, from the DeepMatter™ company InfoChem, is a powerful computer-aided synthesis design tool that does exactly that. It uses machine learning to generate sets of chemical rules (also known as transform libraries) from reaction databases, including proprietary and confidential reaction data in electronic laboratory notebooks (ELNs). In this way, customers can create their own proprietary transform libraries, providing ideas for novel synthetic routes. The end result: more efficient pathways to the required target.

ICSYNTH is run as a web application. The user inputs their target molecule, by simply drawing its structure in the app. They then select their desired synthetic strategy and choose one or more transform libraries to search for matching reactions, using the power of ICFSE, possibly the fastest chemical search engine.

The software will then analyse possible disconnections at every position of the target molecule, with rules built using the industry-leading reaction-centre mapping tool, ICMAP. It automatically generates an interactive, multi-step synthesis tree, on which each node indicates a particular precursor.

The user can control both the size of the tree that is generated, in both breadth (ie the number of possible precursors) and depth (the number of steps generated for each pathway). Each precursor is checked for availability in commercial catalogues – with direct links to preferred suppliers and, where available, cost information – and each reaction step is linked to an analogue in a previously known reaction, whether public or proprietary. By selecting particular nodes the user can view the transformations and save promising pathways, exporting the data in Excel or RDF formats.

ICSYNTH is flexible and customisable, allowing any chemistry data to be integrated and used to generate novel synthesis routes, whether creating entirely new molecules or finding new and more efficient routes to existing ones. It also allows collaborative work across a team. Individual members can focus on particular areas of the synthesis tree, adding comments and exchanging information in the app as they work, increasing efficiency and saving time. As well as allowing users to work with their proprietary reaction data and ELNs, it offers “out of the box” rules based on InfoChem’s extensive database, SPRESI, containing data on millions of molecules and reactions.

While ICSYNTH provides retrosynthetic analysis, working backwards from a target molecule, InfoChem also allows forward reaction planning with the ICFRP add-on, accessed through the same web interface. Given a particular starting material, ICFRP uses the same algorithmic approach to analyse molecules that could be created. Depending on the parameters chosen by the user, this allows reactivity mapping, functional group library design, or scaffold modification. This approach can let the user identify readily achievable modifications with possible active functional groups (bioisosteres), or novel compounds without patent protection. It also predicts potential byproducts (genotoxic impurities or GTIs) that may be created during synthesis.

Future posts will focus in detail on how ICSYNTH and ICFRP can help you optimise your reaction strategies and improve efficiency in your lab.