.
Angewandte
Communications
a cell); in contrast, the network of organic chemistry is
a collection of individual reactions performed by different
chemists at different times (Figure 1b), typically under differ-
ent conditions and in different solvents. Our goal is to find
within NOC the combinations of reactions that can be wired
into the simplest systems, that is, the linear, one-pot sequen-
ces.
consumed before the reagents to make C are added). Filter #2
verifies whether reaction conditions required in each step
would cause unintended reactions of functional groups in
other steps. These rules are summarized in the form of a table
comprising 97 typical reaction types/conditions versus 322
functional groups (Supporting Information, Section 2). Filter
#3 checks for the compatibility of solvents using a digitized
version of the well-known solvent miscibility tables (although
reactions can occur at solvent interfaces, it is not efficient for
one-pot reactions to combine solvents that are not miscible).
Filters #4–#8 are based on the compatibility of 600 common
reagents and are summarized in the Supporting Information,
Section 3. For instance, filter #4 checks for anhydrous versus
aqueous conditions (for example, in Gattermann reactions,
which install aldehyde groups in aryl systems under aqueous
conditions, subsequent one-pot steps cannot involve water-
sensitive reactants or reagents, such as Grignard compounds,
alkali metal hydrides, or organolithium reagents). In a similar
way, filter #5 checks for oxidizing versus reducing conditions,
which are incompatible unless the oxidizing or reducing
reagents can be converted into unreactive spectator species.
Filter #6 determines acid–base compatibility and alerts the
user as to whether acidic, basic, or neutral conditions are
incompatible among reaction steps, such that the addition of
acids/bases at specific times, needs to be planned. Finally,
filter #7 checks for the incompatibilities in terms of hydride/
proton sources, and filter #8 checks for the compatibility of
chemical groups on the reagents (akin to filter #1 for
substrates/products). Although the rules stored in the filter
tables comprise over 86000 chemical criteria to evaluate
candidate one-pot sequences, the entire analysis takes only
a small fraction of a second on a typical desktop computer.
Also, when suitable reactions are identified and potential
conflicts resolved by proper reaction timing (that is, order of
addition), an optional step in the algorithm is to check for the
commercial availability of the substrates and reagents (in the
current version of our software, against the list of ca. 20000
chemicals, mostly from Sigma Aldrich).
To do so, the possible reaction sequences within NOC are
evaluated for one-pot-compatibility by several computational
criteria. The initial, trivial check is whether for a given
sequence (for example, a two-step A!B!C), there already
exists within NOC a direct A!C connection; if so, this
sequence is no longer considered. Assuming no A!C
connection is known, the algorithm applies various filters to
determine the compatibility of the individual reactions to be
wired together. Filter #1 checks for the compatibility of
functional groups on all molecules participating in a putative
sequence. Specifically, a house-written program is first used to
unambiguously partition each of the molecules into functional
groups taken from a list of 322 common chemical function-
alities (Figure 2a; Supporting Information, Section 1). The
constituent groups are then compared against a 322 ꢁ 322
“master” matrix where all possible group combinations are
classified[6] as mutually unreactive (that is, compatible; gray
entries in Figure 2b) or reactive (incompatible; red entries).
If incompatibilities are found, the algorithm suggests the
order of addition to avoid conflicts (for instance, in a two-step
A!B!C sequence, if A reacts with C, then A must be
Of course, the true value of any theoretical–chemical
algorithm is in experimental validation. In principle, the
method can be tested to identify one-pot reactions from
among any of the possible 1.8 billion two-step sequences
present within the NOC. While our algorithm has already
identified over a million (and counting!) possible sequences,
such randomly chosen reactions might be of no real-world
interest, and so herein we chose to illustrate the performance
of the method by “wiring” reaction sequences within classes
of compounds that are of popular interest and/or practical
importance. As the examples in Figure 3 and 4 span 27 one-
pot syntheses (14 two-step, 12 three-step, and 1 four-step
sequences), the main text focuses on their key aspects (full
experimental procedures and structural characterization of all
compounds are included in the Supporting Information,
Section 7). Finally, we emphasize that all the sequences and
yields reported below are based on the one-pot procedures as
suggested by the algorithm; that is, without any human
“tinkering” to optimize the yields, and so on (for comparison
of one-pot versus sequential reactions yields, see the Support-
ing Information, Section 4).
Figure 2. Illustration of the group-compatibility filter (#1). a) Examples
of algorithmic partitioning of molecules into specific functional
groups. The full list of possible 322 groups is included in the
Supporting Information, Section 1. b) A large fragment (left) and
further magnification (right) of the group-compatibility 322ꢀ322
master matrix used to determine the compatibility or incompatibility
of groups involved in a putative one-pot sequence. The classifications
are made assuming typical reaction conditions (see [11]). The zoomed
fragment contains some familiar group combinations and illustrates
their well-known reactivity trends (for example, ethers are poor
nucleophiles and generally unreactive, primary and secondary amines,
on the other hand, are reactive towards all kinds of electrophiles, and
so on).
2
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 1 – 6
These are not the final page numbers!