The Journal of Organic Chemistry
Note
doublet of doublets, ddt = doublet of doublet of triplets, and m =
multiplet. A high-performance liquid chromatography (HPLC)
(Waters Corp., Milford, MA) system was used to quantify the
isomacroin, as product, and the samples were collected from the
chemical reactions. The chromatographic analysis was performed in
Agilent Eclipse XDB-C18 column (150 mm × 4.6 mm i.d. 3.5 μm) at
room temperature. The product was detected by ultraviolet (UV)
absorbance detection at 254 nm. The analyses were performed under
an appropriate gradient of ammonium acetate/acetonitrile (90:10) at
a pH of 8.5 and acetonitrile/ammonium acetate (90:10) and a total
flow rate of 1 mL/min. The injection volume was 5 μL, and the total
run time was 20.1 min. The ultraperformance liquid chromatography
(UPLC) analyses were performed on an ACQUITY UPLC system
equipped with a photodiode array (PDA) detector (Waters Corp.,
Milford, MA) coupled to a mass single-quadrupole detector (QDa
Waters). This system was used to identify the starting raw materials,
product, and byproducts, and the samples were collected from the
chemical reactions. All compounds were monitored at 254 nm by
PDA detector. The single ion recording (SIR) method was established
on a single quadruple mass detector. The QDa conditions were set as
follows: a cone voltage of 15 V, a capillary voltage of 0.8 kV, and a
source temperature of 600 °C. The data were acquired under the SIR
mode. The column, injection volume, flow rate, and solvent managers
were already described previously.
Synthesis of 1-(1-Methyl-1H-imidazol-5-yl)ethan-1-one. To a
flask containing 1-methyl-1H-imidazole (0.485 mL, 0.50 g, 6.0 mmol)
and tetrahydrofuran (2.87 mL) at −78 °C was added 1.6 M n-BuLi in
hexane (4.0 mL, 6.5 mmol) followed by stirring at −78 °C for 40 min.
Then chlorotrimethylsilane (0.8 mL, 6.3 mmol) was added slowly,
and the mixture was stirred at −78 °C for 1 h. The 1.6 M n-BuLi in
hexane (4.0 mL, 6.5 mmol) was added, the cooling bath was removed,
and stirring was continued until the temperature reached 10 °C. The
mixture was recooled to −78 °C, and a solution of N,N-
dimethylacetamide (0.463 mL, 5.0 mmol) was added. The cooling
bath was removed, and the stirring was continued for 40 min at room
temperature. The reaction was quenched with a few drops of
methanol (1.0 mL), and brine was added. The organic layer was
separated, and the aqueous layer was extracted with dichloromethane.
The combined organic phases were dried (sodium sulfate anhydrous),
filtered, and concentrated under reduced pressure. The crude was
filtered through a short silica plug and eluting with ethyl acetate (75
mL). The solvent was evaporated to afford the required compound.
Colorless oil, 93% (0.703 g), Rf 0.68 (1/1 v/v ethyl acetate/heptane).
1H NMR (300 MHz, (CD3)2CO): δ 7.28−7.24 (m, 1H), 7.00 (d, J =
1.0 Hz, 1H), 3.92 (s, 3H), 2.48 (s, 3H). 13C{1H} NMR (75 MHz,
(CD3)2CO): δ 189.7, 128.6, 127.4, 35.2, 26.1. As described in the
literature.47
S5). The reaction rate predictions suggest that formation of 1
occurs rapidly, with the dehydration step being rate-limiting.
The results also indicate that consumption of 2-amino-
benzaldehyde occurs through an aldol condensation toward
intermediate 2. Moreover, 1 is amenable to slow degradation
relative to the elimination step when excess 2-amino-
benzaldehyde and t-BuOK are employed. Domain knowledge
informs that formation of a Schiff base is not viable under basic
conditions. As an adversarial control to the proposed
mechanism, we fitted a model assuming the formation of an
imine as first step. The results are in line with established
intuition and assert the formation of 2 (kimine = 0.01 M−1 s−1 vs
0.05 M−1 s−1), thus providing an additional layer of confidence
on our kinetic model.
On a broader perspective, we confirmed the ML predictions
through an orthogonal means, as both methods in our
investigation converged to identical reaction protocols and
predicted yields. More specifically, kinetic modeling suggests
that a global optimum, with minimization of the byproduct, is
achieved at 78 °C and utilizing 2 and 1.05 molar equiv of 2-
aminobenzaldehyde and t-BuOK, respectively, over 593 min of
reaction (Figure S15). There is, however, value in coalescing
both approaches into a streamlined workflow. The kinetic
model augmented the ML-derived intuition by elucidating the
reaction mechanism and rate constants. It also informed the
impact of the byproduct for the development of an optimized
synthesis protocol, albeit requiring time-consuming computa-
tion (ca. 5−10 min for ML vs 220−250 min for kinetic
model). Together, this substantiates the power of pattern
identification through ML. It also provides a motivation for
employing computationally expensive methods when added
value is warranted.
In conclusion, we implemented a cascaded workflow
comprising adaptive learning and kinetic modeling to facilitate
̈
the access to 1 via Friedlander chemistry and afford physical
chemistry insights. The whole optimization processincluding
featurization, random selection, hyperparameter tuning, and
ML selectionallowed the prioritization of experiments and
identification of an optimum while executing only a minute
amount (3%) of all possible reactions. This is relevant because
ML can equally work as an optimizer or fine-tuner of
experiments, using poor or good yields as starting points,
respectively. Further, the process also allowed establishing a
mechanistic path to the transformation. Our study provides
proof of concept for a viable integration of well-established
concepts in process chemistry with emerging technologies.
Ultimately, it may impact on molecular medicine pipelines by
expediting the access to high value chemical matter for
screening purposes. We foresee this and similar integrations
working as research assistants, promoting probabilistically
informed decisions and democratizing organic syntheses in the
digital chemistry era.
Synthesis of 2-Aminobenzaldehyde. A solution of 2-nitro-
benzaldehyde (0.50 g, 3.3 mmol) in ethanol (9.4 mL) was stirred
for approximately 1 min. Iron powder (0.554 g, 9.9 mmol) and dilute
hydrochloric acid (3.3 mL of 1.0 M HCl, 0.5 mmol) were added to
the stirred solution, and the reaction was heated to reflux for 2 h in an
Asynt DrySyn Single Position Blocks system. The reaction mixture
was cooled to room temperature, diluted with ethyl acetate (27.0
mL), and stirred for 5 min before being filtered through a short Celite
plug. The filtrate was evaporated under reduced pressure to yield a
yellow oil. The product was stored at −20 °C. Yellow oil, 100% (0.40
1
g), Rf 0.44 (1/5 v/v ethyl acetate/hexane). H NMR (300 MHz,
(CD3)2CO) δ 9.88 (d, J = 0.6 Hz, 1H), 7.55 (dd, J = 7.8, 1.6 Hz, 1H),
7.31 (ddd, J = 8.5, 7.0, 1.6 Hz, 1H), 6.81 (dq, J = 8.3, 0.7 Hz, 2H),
6.74−6.66 (m, 1H). 13C{1H} NMR (75 MHz, (CD3)2CO) δ 193.7,
1508, 135.6, 134.9, 118.6, 115.9, 115.4. As described in literature.48
Synthesis of Isomacroin (1). The ketone (0.10 g, 0.80 mmol, 1
molar equiv) and potassium hydroxide (0.057 g, 1.0 mmol, 1.25 molar
equiv) were dissolved in ethanol (6 mL/mmol). Then the 2-
aminobenzaldehyde (0.098g, 0.8 mmol, 1 molar equiv) was added to
the solution and the reaction mixture refluxed 3 h in a Asynt DrySyn
Single Position Blocks system. The solvent was evaporated, and the
crude was purified by flash column chromatography with heptane/
ethyl acetate (3:1) eluent. Yellow solid, 80% (0.135 g), Rf 0.40 (1/1
EXPERIMENTAL SECTION
■
General Methods. Starting materials and reagents were
purchased from Sigma-Aldrich, Alfa Aesar, Fluka, or Acros and used
without further purification. Reactions were carried out on a Radleys
1
Carousel 6 Plus Reaction Station. H NMR spectra were obtained at
on a Bruker Avance 300 MHz in CD3OD and (CD3)2CO with
chemical shift values (δ) in parts per million using residual solvent
peaks as the internal standard, and 13C NMR spectra were obtained at
75 MHz in the same deuterated solvents. Coupling constants (J) are
reported in hertz with the following splitting abbreviations: s = singlet,
d = doublet, t = triplet, dd = doublet of doublets, ddd = doublet of
D
J. Org. Chem. XXXX, XXX, XXX−XXX