Data science enables the development of a new class of chiral phosphoric acid catalysts

The widespread success of BINOL-chiral phosphoric acids (CPAs) has led to the development of several high molecular weight, sterically encumbered variants. Herein, we disclose an alternative, minimalistic chiral phosphoric acid backbone incorporating only a single instance of point chirality. Data science techniques were used to select a diverse training set of catalysts, which were benchmarked against the transfer hydrogenation of an 8-aminoquinoline. Using a univariate classification algorithm and multivariate linear regression, the key catalyst features necessary for achieving high levels of selectivity were deconvoluted, revealing a simple catalyst model capable of predicting selectivity for out-of-set catalysts. This workflow enabled extrapolation to a catalyst that provided higher selectivity than both peptide-type and BINOL-type catalysts reported previously (up to 95:5 er). These techniques were then successfully applied toward two additional transforms. Taken together, these examples illustrate the power of combining rational design with data science (ab initio) to efficiently explore reactivity during catalyst development.