Mapping Alkyne Chemical Space—Classifying the Reactivity of Alkynes in Semihydrogenation and beyond Using a Data Science Framework

Abstract

A data science framework that integrates DFT featurization, dimensionality reduction, and hierarchical clustering is employed to study the chemical space of commercially available alkynes, highlighting the differentiation of over 2700 structures into 6 subsets. Referencing the substrate scope of a recent publication, the selected molecules are shown to be representative of the alkyne chemical space by covering each of the 6 clusters. Logistic regression classification allows the categorization of the tested alkynes according to their propensity to hydrogenate under milder or harsher reaction conditions. This categorization identified highly chemically shielded alkynes with low total dipole as the easiest to hydrogenate, likely by breaking the cylindrical symmetry of the alkyne. In summary, we identify potential reasons for the difference in reactivity of alkyne substrates in Ni₃Zn nanocrystal catalyzed semi-hydrogenation and present a useful workflow and alkyne data set for studying catalytic reactions in general.