From building block identification to reaction space exploration: novel tools to make the most out of chemical reactions
Nikolaus Stiefl, Nadine Schneider, Daniel M. Lowe, Roger A. Sayle, Michael A. Tarselli, Gregory A. Landrum
Even though chemical reactions form the basis of all medicinal chemistry projects in pharmaceutical research, only few public data sources are available to develop new methods around handling all the different nuances related to chemical reactions. Especially public data sources containing reaction schemes in a machine-readable format are still quite rare.
Based on data recently obtained from a massive text-mining [1, 2] effort on over a million US patents we have developed several new tools to explore chemical reactions. The methods target both, reaction curation as well as analysis [3, 4]. Here we will present an overview of the algorithms with a main focus on reaction data preprocessing and reaction classification. Special attention will be given to the reaction role assignment procedure which is based on molecular fingerprints and does not rely on a maximum common substructure .
Finally, we will highlight how we make use of the information obtained with an example focusing on building block analysis.