Synthetic route design assessment using vectors derived from similarity and complexity

With the aim of improving the machine-interpretation of synthetic routes we describe a new theoretical approach to visualising and assessing synthetic pathways in the absence of empirical data such as yield, cost and waste. The representation of molecular structures as coordinates derived from molecular (fingerprint) similarity and complexity allows individual transformations to be viewed as vectors (reactant to product) whereby the magnitude and direction of travel can be used to assess and quantify transformation efficiency. Vectors derived in this way are shown to follow logical trends when grouped by reaction type/class. Synthetic routes can thus be visualised as a series of head-to-tail vectors (one per transformation or step) traversing the range between starting material and target whereby the efficiency with which this range is covered can be quantified. Our approach is built upon the analysis of > 350k literature syntheses (> 1.4m reactions), is readily machine-interpretable and avoids the challenges associated with automated reaction class assignment and atom-mapping.