Transform Descriptors for template free reaction classification

Since the explosion of the artificial intelligence-based methods applied to organic synthesis, many research groups have been scouting different possibilities to describe organic reactions incl predicted outcome using various descriptor formats. From fingerprints to quantum mechanics descriptors these have been used to predict yields, product structures, catalysts, solvents, and reagents. Though promising results have been published from different research groups, the general applicability, as well as the difficulty in understanding the descriptors used makes these approaches far too complicated for many experimental synthetic chemists. In this article, we present a simple to understand and simple to calculate set of transform descriptors (TDs) capable of mimicking the results of those methods using much more complex approaches. These transform descriptors have been successfully used for reaction classification and cluster prediction of a hypothetical newly designed reaction.