developed by Antilla and co-workers ( 41 )was
selected for several reasons. The reaction is
high yielding and highly reproducible; it can
be performed under air at room temperature,
thus facilitating rapid screening; and a range
of selectivities (0 to 99%) has been reported
with different catalysts (six reported catalysts).
Accordingly, we judged this reaction to be a
good candidate for empirically evaluating the
selectivity space covered in the UTS. By calcu-
lating ASO and electronic descriptors for reac-
tants and products as well and concatenating
these descriptors with catalyst descriptors, indi-
vidual reaction profiles could be constructed
that also took into account substrate properties.
The inclusion of substrate descriptors also
increased the number of data points obtained
per catalyst synthesized. As a general note on
the use of reactant and product descriptors, we
find that it provides the following benefits:
More data points can be collected per catalyst
Zahrtet al.,Science 363 , eaau5631 (2019) 18 January 2019 5of11
Fig. 4. BINOL phosphoric acid (BPA) UTS.(A) UTS of phosphoric acid catalysts selected by the Kennard-Stone algorithm. (B) Average selectivity of
training catalysts across the 16 training reactions. Ph, phenyl; Me, methyl; Et, ethyl.
RESEARCH | RESEARCH ARTICLE
on January 18, 2019^
http://science.sciencemag.org/
Downloaded from