(595k) Human-Interpretable Reaction Informatics | AIChE

(595k) Human-Interpretable Reaction Informatics

Cognitive technologies in the areas dominated by expert knowledge, such as chemical and material sciences, require cheminformatic models that are human-interpretable and can be mapped onto the existing structure of the domain knowledge. Advanced analysis of chemical data associated with molecules and reactions relies on fingerprinting methods that enable direct comparison of chemical entities with each other. We will present a new strategy of reaction fingerprinting that meets requirements of human readability and relates to the chemical intuition better than the strategies explored so far. Its application to different classes of reactions will be discussed.