Exploration of flow reaction conditions using machine-learning for enantioselective organocatalyzed Rauhut-Currier and [3+2] annulation sequence

Article abstract of DOI:10.1039/c9cc08526b

A highly atom-economical enantioselective organocatalyzed Rauhut-Currier and [3+2] annulation sequence has been established by using a flow system. Suitable flow conditions were explored through reaction screening of multiple parameters using machine learning. Eventually, functionalized chiral spirooxindole analogues were obtained in high yield with good ee as a single diastereomer within one minute.

Full text of DOI:10.1039/c9cc08526b

View Article Online
View Journal
ChemComm
Chemical Communications
Accepted Manuscript
This article can be cited before page numbers have been issued, to do this please use: M. Kondo, H. D. P.
Wathsala, M. Sako, Y. Hanatani, K. Ishikawa, S. Hara, T. Takaai, T. Washio, S. Takizawa and H. Sasai, Chem.
Commun., 2019, DOI: 10.1039/C9CC08526B.
Volume 54
This is an Accepted Manuscript, which has been through the
Number
January 2018
Pages 1-112
1
4
Royal Society of Chemistry peer review process and has been
accepted for publication.
ChemComm
Chemical Communications
rsc.li/chemcomm
Accepted Manuscripts are published online shortly after acceptance,
before technical editing, formatting and proof reading. Using this free
service, authors can make their results available to the community, in
citable form, before we publish the edited article. We will replace this
Accepted Manuscript with the edited and formatted Advance Article as
soon as it is available.
You can find more information about Accepted Manuscripts in the
Information for Authors.
Please note that technical editing may introduce minor changes to the
text and/or graphics, which may alter content. The journal’s standard
Terms & Conditions and the Ethical guidelines still apply. In no event
shall the Royal Society of Chemistry be held responsible for any errors
or omissions in this Accepted Manuscript or any consequences arising
from the use of any information it contains.
ISSN 1359-7345
COMMUNICATION
S. J. Connon, M. O. Senge et al.
Conformational control of nonplanar free base porphyrins:
towards bifunctional catalysts of tunable basicity
rsc.li/chemcomm

Please do not adjust margins
Page 1 of 5
ChemComm
View Article Online
DOI: 10.1039/C9CC08526B
COMMUNICATION
Exploration of flow reaction conditions using machine-learning for
enantioselective organocatalyzed Rauhut-Currier and [3+2]
annulation sequence
Received 00th January 20xx,
Accepted 00th January 20xx
Masaru Kondoa, H.D.P. Wathsalaa, Makoto Sakoa, Yutaro Hanatania, Kazunori Ishikawab, Satoshi
Harab, Takayuki Takaaib, Takashi Washio*b,c, Shinobu Takizawa*a,c, Hiroaki Sasai*a
DOI: 10.1039/x0xx00000x
Exploration of Suitable Reaction Conditions
A
highly atom-economical enantioselective organocatalyzed
a) Conventional Method
b) This Work
Rauhut-Currier and [3+2] annulation sequence has been
established by using flow system. The suitable flow conditions were
explored through reaction screening of multiple parameters using
machine learning. Eventually, functionalized chiral spirooxindole
analogues were obtained in high yield with good ee as a single
diastereomer within one minute.
screening of each factor
machine learning (ML)-assisted method

+
ML
trial and error performing
many reactions (inefficient)
minimum
reactions
machine learning
(Gaussian process regression)
High Efficiency
 save cost
 save time
 save chemicals
 save manpower
 reduce waste
In the field of organic chemistry, the optimization of conditions is a
crucial and unavoidable process in the development of a synthetic
reaction in both academic and industrial research. The conventional
procedure is to vary the reaction parameters individually while
keeping the others constant (Fig. 1a). Therefore, an efficient and
rapid reaction screening strategy is of significant interest to the
scientific community. Machine-Learning (ML) is a robust and reliable
tool that can be used to achieve efficient optimization.1 Over the past
decade, ML has been applied to various chemical fields such as drug
discovery,2 synthesis planning,3 and material and catalyst design.4
More recently, outstanding results with automated5 and
computational optimization procedures6 have been reported. In
theoretical and physical chemistry, Gaussian process regression
(GPR),7 which is a kernel-based statistical learning algorithm, has
been increasingly applied to predict a variety of chemical properties
as a black-box estimator under a given up to date dataset. Since it is
difficult for chemists to understand the unclear tendency and
correlation of each parameter of a novel reaction and its outcome
without performing a thorough reaction optimization, GPR could be
helpful in estimating the optimal conditions from experimental
results by using a kernel-based method and conducting the minimum
number of reactions for multi-parameter screening (e.g. flow
reaction and electrolysis). (Fig. 1b).
Fig. 1 Exploration of suitable reaction conditions through minimum
experimental data with machine learning.
strategies capable of constructing multiple chiral centers are still in
high demand.8a-c,9 As part of our exploration of enantioselective
domino processes with a single operation,10 we were interested in
developing the organocatalyzed Rauhut-Currier (RC) and [3+2]
annulation sequence11 of dienone 1 with allenoate 2 to access the
highly functionalized chiral spirooxindole analogue 4, bearing three
contiguous chiral centers. The domino reaction of 1a and 2a as
prototypical substrates in the presence of chiral organocatalyst 5 (20
mol %) was initially attempted in a batch process (Table 1). Among
the organocatalysts examined, (S)-valine-derived catalyst 5a bearing
diphenylphosphine and benzoyl units promoted the desired domino
reaction to afford spiro heterocycle 4a with high regio- and
stereoselectivity, albeit in 42% yield. In 2017, we and Huang
independently reported highly active organocatalyst for the
enantioselective intramolecular RC reaction of 1,10c,11d however,
neither amine catalyst 5e nor Huang Cat. for the RC showed
attractive outcomes on our desired domino reaction. Many
unidentified products were also formed due to side reactions of
dienone 1a with the highly reactive intermediary RC product 3a,10c
and allenoate 2a in the presence of catalysts 5. In terms of
enantioselectivity, the use of catalyst 5a in toluene gave 4a in 92%
ee as a single diastereomer, but no significant improvement in the
chemical yield of 4a was accomplished in the batch system (see
Supplementary Table S1, S2).
Spirooxindole motifs are found in numerous natural products and
biologically active molecules such as horsfiline, gelsemine, and
marcfortine B.8 To date, spirooxindole derivatives have attracted
much attention in the area of antiviral drug discovery and
development, owing to the high number of positive hits achieved by
this scaffold.8d-f Although significant progress has been made in the
When we considered the reaction processes, we found that a micro-
mixing flow system12 offered a means of improving the chemical
yield of product 4a by suppressing undesired side reactions. Under
asymmetric synthesis of diverse spirooxindoles, facile synthetic
a The Institute of Scientific and Industrial Research (ISIR), Osaka University,
Mihogaoka, Ibaraki, Osaka 567-0047, Japan
E-mail: taki@sanken.osaka-u.ac.jp, sasai@sanken.osaka-u.ac.jp
b Department of Reasoning for Intelligence, ISIR, Osaka University,
E-mail: washio@ar.sanken.osaka-u.ac.jp
rapid mixing with
a
micro mixer: Comet X-0113 at dilute
concentration of a mixture of substrates 1a and 2a, and catalyst 5a
(0.01 M for 1a in toluene), the intramolecular RC reaction initially
proceeded to afford
c Artificial Intelligence Research Center, ISIR, Osaka University
Electronic
Supplementary
Information
(ESI)
available:
See
DOI: 10.1039/x0xx00000x
Please do not adjust margins

Please do not adjust margins
ChemComm
Page 2 of 5
COMMUNICATION
Journal Name
Table 2. Exploration of suitable reaction conditionsa
PPh2
Table 1. Catalyst screening in batch system
View Article Online
DOI: 10.1039/C9CC08526B
O
NHBz
(S)- (20 mol%)
O
5a
full conversion
O
CO2Et
CO2Et
φ = 1 mm
l = 1 m
5
chiral organocatalyst (20 mol %)
H
O
toluene
Flow rate
CO2Et
CO2Et
micro mixer
H
Me
Me
TsN
toluene (0.1 M), 0.5 h, r.t.
TsN
toluene
O
2a
(2 equiv.)
Me
N
toluene (0.01 M), Temp.
O
1a
4a
Ts
O
4a
+
•
89-94% eeb
Me
Ts
N
O
Ph2P
NHBz
Ph2P
NHBz
Ph2P
NHAc
Ph2P
NHTf
1a
2a
(X equiv.)
5a
5b
5c
5d
42% yield
86% ee
34% yield
82% ee
trace
32% yield
48% ee
Flow rate
(mL/min)
2a
(equiv.)
NMR yieldc
(%)
Entry
Temp. (°C)
O
Ph2P HN
1
2
1.0
1.5
2.0
2.5
3.0
1.7
1.7
1.7
1.7
1.7
1.7
90
80
2.0
49
72
Et2N
NHBz
Et2N
NHBz
Huang Cat.
31% yield
64% ee
5e
5f
0%a
HO
2.0
0%a
3
70
2.0
58
1a
2a 5
(0.06 mmol), and catalyst (S)- (20 mol%), in dry
toluene. Yields were estimated by 1H NMR using 1,3,5-trimethoxybenzene as an internal
Reaction conditions:
(0.03 mmol),
4
60
2.0
55
3a
standard. Ees were determined by HPLC analysis (Daicel Chiralpak ID). [a] RC product
was only observed.
5
100
80
2.0
43
the highly reactive dienone RC product 3a, followed by
intermolecular [3+2] annulation, leading to 4a in 49% NMR yield
(Table 2, entry 1).14 This improved outcome prompted us to optimize
reaction conditions in the flow system. Moreover, GPR on GPy, which
is a programming library for Gaussian processes regression,15
constructs a regression model by using a limited number of observed
data through ML, and searches a subsequent appropriate parameter
value by using the model as a surrogate model of the flow reaction
conditions. To achieve the ML estimation of the suitable flow
reaction conditions, the temperature (60–100 °C) and flow rate (1.0–
3.0 mL/min) were first screened using GPy while maintaining the
quantity of 2a at 2.0 equiv. to minimize the cost of chemicals (entries
1–5) (see Supplementary Table S3 and Fig. S3)16 with referencing the
batch optimization outcomes. Standardization was performed to
compensate for the large difference in scale between temperature
and flow rate. Evaluation of the observed results using GPR
illustrated the estimated yield from screening results of flow rate and
temperature as shown in Fig. 2A (intense yellow: higher yield; intense
black: lower yield).
6
1.0
45
7
100
90
1.5
51
8
2.0
58
9
70
3.0
50
10
11d
60
4.0
48
80
2.0
78 (76)e
aReaction conditions: 1a (0.03 mmol), 2a and catalyst (S)-5a (20 mol%), in
degassed dry toluene (3 mL), micro mixer: Comet X-01, stainless-steel tube
(internal diameter: 1.0 mm, length: 1.0 m) bEnantiomeric excess was
determined by HPLC analysis (Daicel Chiralpak ID). c1,3,5-
Trimethoxybenzene was used as an internal standard. d4a was obtained in
94% ee. eIsolated yield.
This outcome encouraged us to focus on the yellow area in Fig. 2A
predicting the optimal conditions. The confidence bounds (light blue
area) in Fig. 2B and 2C indicated the flow rate and temperature that
should be used to conduct the next experiments. The optimal flow
rate and temperature were estimated to be 1.7 mL/min and 77 °C
from the expected maximum value of the upper confidence bounds
in Fig. 2B and 2C (red line). Secondly, the yield through screening of
temperature (60–100 °C) and quantity of 2a (1.0–4.0 equiv.) was
estimated from Table 2 (entries 6–10) using a fixed flow rate of 1.7
mL/min as shown in Fig. 2D, Fig. 2E, and 2F (see Supplementary Table
S4 and Fig. S4) show the next evaluations of temperature and the
amount of 2a obtained from the yellow ring in Fig. 2D. On the basis
of Fig. 2E and 2F, the ML predicted reaction conditions were 2.0
equiv. of 2a with a reaction temperature below 90 °C. Our ML-
assisted exploring methodology of reaction conditions could narrow
down multi-parameters more efficiently in comparison of the
conventional trial and error screening. Considering all the results
estimated by GPR and the experimental data in batch system, we
decided the suitable flow reaction conditions as 2.0 equiv. of 2a and
a flow rate of 1.7 mL/min at 80 °C. In fact, when we applied the
reaction conditions to a practical reaction, 4a was isolated in 76%
yield and 94% ee (Table 2, entry 11).17
(A)
(D)
(B)
(E)
(F)
(C)
Fig. 2 Gaussian process regression with GPy. (A) Estimated yield from
Table 2 (entries 1–5); (B) Predicted yield for flow rates in the yellow ring
in Fig. 2A; (C) Predicted yield for temperatures in the yellow ring in Fig.
2A; (D) Estimated yield from Table 2 (entries 6–10); (E) Predicted yield for
equivalents of 2a in the yellow ring in Fig. 2D; (F) Predicted yield for
temperatures in the yellow ring in Fig. 2D.
Having the estimated optimal conditions, we investigated the
substrate scope of the domino reaction (Scheme 1). Compounds 1b
(R1 = Et) and 1c (R1 = iPr) with aliphatic substituents afforded products
2 | J. Name., 2012, 00, 1-3
This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins

Page 3 of 5
ChemComm
Please do not adjust margins
Journal Name
COMMUNICATION
products 4o (45% yield, 92% ee) and 4p (54% yield,V9ie2w%Arteicele).OnTlhinee
chiral spirocycles 4q and 4r were obtained wDitOhIi:n1201.10s3r9e/sCid9eCnCc0e85ti2m6Be
in 89% yield and 88% ee, and 76% yield and 86% ee, respectively
(flow rate: 1.5 mL/min). Similarly, dienone 1e and benzyl allenoate
2b were also converted into the corresponding product 4s in 73%
yield and 90% ee under the same conditions as for 4q and 4r. The
absolute configuration of chiral spiro compound 4e with catalyst (S)-
5a was determined to be S,R,R by single X-ray crystallographic
analysis (see Supplementary Fig. S6).
Scheme 1. Scope of RC and [3+2] annulation sequence.
5a
(S)- (20 mol %)
O
full conversion
CO2R3
φ = 1 mm
l = 1 m
1.7 mL/min
micro mixer
H
R1
O
N
toluene (0.01 M)
80 °C, residence time <26 s
R2
CO2R3
O
4
+
•
(single diastereomer)
R1
N
1
R2
2
(2.0 equiv)
O
O
O
O
A plausible reaction mechanism for this sequential reaction is shown
in Fig. 3. Under the dilute concentration of substrates, initially, the
intramolecular RC reaction proceeds as: The addition of catalyst 5a
to the acrylamide terminal on 1a generates Intermediate A (Int. A).
The chiral enolate in Int. A. subsequently reacts with one of the
olefins on the dienone unit to generate Int. B. Then, deprotonation
of α-proton of the carbonyl group in Int. B affords Int. C (3a) bearing
a highly reactive exo-olefin, which is subsequently trapped with
allenoate 2a via [3+2] annulation. In the intermolecular [3+2]
annulation supported by the micro-mixing flow system, allenoate 2a
reacts with regenerated organocatalyst 5a to afford Int. D, resulting
in Michael reaction of the exo-olefin in Int. C, despite the dilute
concentration of the reactants (significantly important, to suppress
side reactions in the presence of many highly reactive species as
shown in Fig. 3). Next, the electron deficient phosphonium cation in
Int. E prompts the conjugate addition of enolate to give Int. F. Finally,
the desired product 4a is formed through proton-transfer and
CO2Et
CO2Et
CO2Et
H
H
H
N
N
N
Ts
O
4b
Ts
O
4c
Ts
O
4d
68% yield, 92% ee
92% yield, 96% ee
61% yield, 91% ee
O
O
O
CO2Et
CO2Et
CO2Et
H
O
H
H
O
MeO
N
N
Ts
4f
N
Ts
O
Ts
4e
4g
78% yield, 94% ee
54% yield, 92% ee
75% yield, 92% ee
O
O
O
CO2Et
F
CO2Et
CO2Et
H
H
O
H
O
N
Ts
4h
N
N
tBu
O
Ts
Cl
Ts
4i
4j
70% yield, 98% ee
81% yield, 87% ee
85% yield, 87% ee
O
O
O
CO2Et
CO2Et
CO2Et
H
H
H
elimination of catalyst from Int. F.
Me
N
Ts
4k
N
Ts
4l
N
Ph
O
Ms
CO2Et
P*
H
O
O
O
4m
H
1a
58% yield, 93% ee
O
60% yield, 96% ee
O
59% yield, 88% ee
O
4a
Me
N
Ts
O
O
CO2Bn
CO2Et
CO2E
t
Int.
F
H
P*
O
H
H
O
Ph2
P
NHBz
5a
Me
N
CO2Et
P*
Me
Intramolecular
Rauhut-Currier
Cycle
Intermolecular
[3+2] Annulation
Cycle
H
MeO
(S)-
= P*
N
N
Me
N
S
O
Ts
Ts
Int.
O
Ts
O
O
Me
N
O
A
O
Ts
4n
2a Int.
+
C (3a)
4o
4p
Int. E
64% yield, 94% ee
45% yield, 92% ee[a]
54% yield, 92% ee[a]
O
O
O
O
O
CO2Et
P*
P*
H
Int. C
(3a)
H
H
CO2Et
CO2Et
CO2Bn
H
H
O
H
Me
Ts
N
Int. D
Me
O
N
Ts
Int. B
O
N
N
N
Int. C (3a)
Ms
Br
Ts
Ts
O
O
Fig. 3 Proposed reaction mechanism.
4q
4r
76% yield, 86% ee[b]
4s
89% yield, 88% ee[b]
73% yield, 90% ee[b]
Yields are those of isolated 4. The ee value of 4 was determined by HPLC
analysis (Daicel Chiralpak ID for 4b-4e, 4g, 4h, 4k and 4p; Daicel Chiralpak IC
for 4f, 4i, 4j, 4n, 4r and 4s; Daicel Chiralpak IB for 4m and 4q; and Daicel
Chiralpak IE for 4l and 4o). Reaction conditions: 1 (0.03 mmol), 2 (2.0 equiv.),
and catalyst (S)-5a (20 mol%) in degassed dry toluene (3 mL), micro mixer:
Comet X-01, stainless-steel tube: internal diameter: 1.0 mm; length: 1.0 m,
flow rate: 1.7 mL/min, residence time: 26 s, at 80 °C. a1.5 equiv. of 2a.
bResidence time: 21 s (flow rate: 2.1 mL/min).
We demonstrated the highly atom-economical chemo-, regio-,
enantio-, and diastereoselective domino reaction initiated by RC and
[3+2] annulation sequence using an organocatalyst in a flow system.
GPR was successfully applied to multi-parameter reaction screening.
The present domino reaction provided chiral spiro heterocycles with
three contiguous chiral centers in excellent results within one minute
under the optimal conditions. This helpful exploration methodology
of suitable reaction conditions will be applicable to other time-
consuming reaction optimizations. Further practical applications of
GPR for other reaction optimization as well as the use of an
immobilized catalyst are also under way in our laboratory.
4b (68% yield, 92% ee) and 4c (92% yield, 96% ee). Similarly, starting
material 1d bearing a vinyl group provided the corresponding
spirocyclic compound 4d in 61% yield and 91% ee. Dienones 1e–l
bearing an electron-donating or electron-withdrawing substituent
on the aryl group reacted with allenoate 2a to give the corresponding
products 4e–l (54–85% yield, 87–98% ee). Additionally, the acrylic
amide containing a methanesulfonyl group as the R2 group (1m)
yielded spiro heterocycle 4m with moderate success (59% yield, 88%
ee). The reaction of dienone 1a with benzyl allenoate 2b could be
achieved to give product 4n in 64% yield and 94% ee. The reaction of
1o (R1 = ethynyl, R2 = tosyl) and 1p (R1 = methyl, R2 = 4-
methoxyphenyl sulfonyl) with 2a (1.5 equiv.) provided desired
Conflicts of interest
There are no conflicts of interest to declare.
Notes and references
1. (a) J. Bergstra, R. Badenet, Y. Bengio and B. Kégl, NIPS, 2011, 24,
2546−2554; (b) J. Bergstra and Y. Bengio, J. Mach. Learn. Res.,
2012, 13, 281−305; (c) T. Back, Evolutionary algorithms in theory
This journal is © The Royal Society of Chemistry 20xx
J. Name., 2013, 00, 1-3 | 3
Please do not adjust margins

Please do not adjust margins
ChemComm
Page 4 of 5
COMMUNICATION
Journal Name
and practice: Evolution strategies, evolutionary programming,
genetic algorithms, Oxford University Press, New York, 1996.
2016, 2, 382–392; (f) S. S. Panda, R. A. Jones, P. Bachawala and P.
2. (a) G. Schneider, Nat. Rev. Drug Discov., 2018, 17, 97–113; (b) 9. For selected recent reviews and reports DonOIa:s1y0m.10m3e9/tCri9cCsCyn0t8h5e2s6iBs
View Article Online
P. Mohapatra, Mini-Rev. Med. Chem., 2017, 17, 1515–1536.
H. Chen, O. Engkvist, Y. Wang, M. Olivecrona and T. Blaschke,
Drug Discov. Today, 2018, 23, 1241–1250; (c) X. Yang, Y. Wang,
R. Byrne, G. Schneider and S. Yang, Chem. Rev., 2019, 119, 10520-
10594.
of spirooxindoles, see: (a) G.-J. Mei and F. Shi, Chem. Commun.,
2018, 54, 6607–6621; (b) S. Takizawa, K. Kishi, M. Kusaba, B.
Jianfei, T. Suzuki and H. Sasai, Heterocycles, 2017, 95, 761–767;
(c) W. He, J. Hu, P. Wang, L. Chen, K. Ji, S. Yang, Y. Li, Z. Xie and
W. Xie, Angew. Chem., Int. Ed., 2018, 57, 3806–3809; (d) T. Kang,
P. Zhao, J. Yang, L. Lin, X. Feng and X. Liu, Chem. Eur. J., 2018, 24,
3703–3706.
3. (a) S. Szymkuć, E. P. Gajewska, T. Klucznik, K. Molga, P. Dittwald,
M. Startek, M. Bajczyk and B. A. Grzybowski, Angew. Chem., Int.
Ed., 2016, 55, 5904−5937; (b) J. N. Wei, D. Duvenaud and A.
Aspuru-Guzik, ACS Cent. Sci., 2016, 2, 725−732; (c) C. W. Coley, 10. Our recent reports on domino reaction, see: (a) S. Takizawa, K.
R. Barzilay, T. S. Jaakkola, W. H. Green and K. F. Jensen, ACS Cent.
Sci., 2017, 3, 434−443; (d) M. H. S. Segler, M. Preuss and M. P.
Waller, Nature, 2018, 555, 604–610; (e) C. W. Coley, W. H. Green
and K. F. Jensen, Acc. Chem. Res., 2018, 51, 1281–1289; (f) O.
Engkvist, P.-O. Norrby, N. Selmi, Y.-H. Lam, Z. Peng, E. C. Sherer,
W. Amberg, T. Erhard and L. A. Smyth, Drug Discov. Today, 2018,
23, 1203−1218; (g) A. F. de Almeida, R. Moreira and T. Rodrigues,
Kishi, Y. Yoshida, S. Mader, F. A. Arteaga, S. Lee, M. Hoshino, M.
Rueping, M. Fujita and H. Sasai, Angew. Chem., Int. Ed., 2015, 54,
15511–15515; (b) M. Sako, Y. Takeuchi, T. Tsujihara, J. Kodera, T.
Kawano, S. Takizawa and H. Sasai, J. Am. Chem. Soc., 2016, 138,
11481–11484; (c) K. Kishi, F. A. Arteaga, S. Takizawa and H. Sasai,
Chem. Commun., 2017, 53, 7724–7727; (d) K. Kishi, S. Takizawa
and H. Sasai, ACS Catal., 2018, 8, 5228–5232.
Nat. Rev. Chem., 2019, 3, 589–604; (h) J. S. Schreck, C. W. Coley 11. For selected reviews and reports on enantioselective
and K. J. M. Bishop, ACS Cent. Sci., 2019, 5, 970–981.
4. (a) P. Raccuglia, K. C. Elbert, P. D. F. Adler, C. Falk, M. B. Wenny,
A. Mollo, M. Zeller, S. A. Friedler, J. Schrier and A. J. Norquist,
Nature, 2016, 533, 73–76. (b) M. S. Sigman, K. C. Harper, E. N.
Bess and A. Millo, Acc. Chem. Res., 2016, 49, 1292–1301. (c) B.
Sanchez-Lengeling and A. Aspuru-Guzik, Science, 2018, 361, 360–
365. (d) K. T. Butler, D. W. Davies, H. Cartwright, O. Isayev and A.
Walsh, Nature, 2018, 559, 547–555. (e) A. F. Zahrt, J. J. Henle, B.
T. Rose, Y. Wang, W. T. Darrow and S. E. Denmark, Science, 2019,
363, eaau5631. (f) D. J. Durand and N. Fey, Chem. Rev., 2019, 119,
6561–6594.
organocatalyzed domino reactions, see: (a) C. Zhang and X. Lu. J.
Org. Chem., 1995, 60, 2906–2908; (b) Domino reactions:
Concepts for efficient organic synthesis (Ed.: Tietze, L. F.), Wiley-
VCH, Weinheim, 2014; (c) P. Chauhan, S. Mahajan and D. Enders,
Acc. Chem. Res., 2017, 50, 2809–2821; (d) H. Jin, Q. Zhang, E. Li,
P. Jia, N. Li and Y. Huang, Org. Biomol. Chem., 2017, 15, 7097–
7101; (e) X.-Y. Chen, S. Li, F. Vetica, M. Kumar and D. Enders,
iScience, 2018, 2, 1–26; (f) X. Wu, L. Zhou, R. Maiti, C. Mou, L. Pan
and Y. R. Chi, Angew. Chem., Int. Ed., 2019, 58, 477–481; (g) H.
Wang, J. Zhang, Y. Tu and J. Zhang, Angew. Chem., Int. Ed., 2019,
58, 5422–5426; (h) K. Li, T. P. Gonçalves, K.-W. Huang and Y. Lu,
Angew. Chem., Int. Ed., 2019, 58, 5427–5431.
5. (a) B. J. Reizman and K. F. Jensen, Acc. Chem. Res., 2016, 49,
1786−1796. (b) V. Sans and L. Cronin, Chem. Soc. Rev., 2016, 45, 12. For selected publications on flow reactions, see: (a) J. Wegner, S.
2032–2043. (c) D. E. Fitzpatrick, T. Maujean, A. C. Evans and S. V.
Ley, Angew. Chem., Int. Ed., 2018, 57, 15128–15132. (d) A.-C.
Bédard, A. Adamo, K. C. Aroh, M. G. Russell, A. A. Bedermann, J.
Torosian, B. Yue, K. F. Jensen and T. F. Jamison, Science, 2018,
361, 1220–1225. (e) M. Trobe and M. D. Burke, Angew. Chem.,
Int. Ed., 2018, 57, 4192–4214.
Ceylan and A. Kirschning, Adv. Synth. Catal., 2012, 354, 17–57;
(b) I. R. Baxendale, J. Chem. Technol. Biotechnol., 2013, 88, 519–
552; (c) J. C. Pastre, D. L. Browne and S. V. Ley, Chem. Soc. Rev.,
2013, 42, 8849–8869; (d) T. Tsubogo, T. Ishiwata and S.
Kobayashi, Angew. Chem., Int. Ed., 2013, 52, 6590–6604; (e) A. A.
Folguerias-Amador and T. Wirth, J. Flow Chem., 2017, 7, 94–95;
(f) J. Britton and C. L. Raston, Chem. Soc. Rev., 2017, 46, 1250–
1271; (g) M. B. Plutschack, B. Pieber, K. Gilmore and P. H.
Seeberger, Chem. Rev., 2017, 117, 11796–11893; (h) R. Gérardy,
N. Emmanuel, T. Toupy, V.-E. Kassin, N. N. Tshibalonza, M.
Schmitz and J.-C. M. Monbaliu, Eur. J. Org. Chem., 2018, 2301–
2351; (i) S. Santoro, F. Ferlin, L. Ackermann and L. Vaccaro,
Chem. Soc. Rev., 2019, 48, 2767–2782. (j) D. Perera, J. W. Tucker,
S. Brahmbhatt, C. J. Helal, A. Chong, W. Farrell, P. Richardson, N.
W. Sach, Science, 2019, 359, 429-434. (k) C. W. Coley, D. A.
Thomas III, J. A. M. Lummiss, J. N. Jaworski, C. P. Breen, V. Schultz,
T. Hart, J. S. Fishman, L, Rogers, H. Gao, R. W. Hicklin, P. P.
Plehiers, J. Byington, J. S. Piotti, W. H. Green, A. J. Hart, T. F.
Jamison, K. F. Jensen, Science, DOI: 10.1126/science.aax1566.
6. (a) Z. Zhou, X. Li and R. N. Zare, ACS Cent. Sci., 2017, 3,
1337−1344.
(b) H. Gao, T. J. Struble, C. W. Coley, Y. Wang, W. H. Green and K.
F. Jensen, ACS Cent. Sci., 2018, 4, 1465-1476. (c) L. Ḧase, L. M.
Roch, C. Kreisbeck and A. Aspuru-Guzik, ACS Cent. Sci., 2018, 4,
1134–1145. (d) A. M. Schweidtmann, A. D. Clayton, N. Holmes, E.
Bradford, R. A. Bourne and A. Lapkin, Chem. Eng. J., 2018, 352,
277−282. (e) P. Vámosi, K. Matsuo, T. Masuda, K. Sato, T. Narumi,
K. Takeda and N. Mase, Chem. Rec., 2019, 19, 77–84.
7. For selected recent reports on GPR for theoritical- and physical
chemistry, see: (a) C. E. Rasmussen and C. K. Williams, Gaussian
processes for machine learning; MIT Press: Cambridge, MA, USA,
2006; Vol. 1. (b) J. P. Aborzpour, D. P. Tew and S. Habershon, J.
Chem. Phys., 2016, 145, 174112. (c) O.-P. Koistinen, F. B. 13. A micromixing device, ‘Comet X-01’, is available from Techno
Dagbjartsdóttir, V. Ásgeirsson, A. Vehtari and H. Jónsson, J.
Chem. Phys., 2017, 147, 152720. (d) R. Ramakrishnan and O. A.
Applications Co., Ltd., 34-16-204, Hon, Denenchofu, Oota, Tokyo,
145-0072, Japan.
von Lilienfeld, Reviews in computational chemistry, Vol. 30; John 14. RC reaction of dienone 1a, followed by introduction of allenoate
Wiley & Sons, 2017; pp 225−256. (e) A. Denzel and J. Kästner, J. 2a in flow system decreased yield of product 4a.
Chem. Theory Comput., 2018, 14, 5777–5786. (f) G. Schmitz and 15. The GPy authors. GPy: A Gaussian process framework in python.
O. Christiansen, J. Chem. Phys., 2018, 148, 241704; (g) G. http://github.com/ SheffieldML/Gpy.
Schmitz, D. G. Artiukhin and O. Christiansen, J. Chem. Phys., 2019, 16. Although three parameters (flow rate, temperature, and
150, 131102.
equivalent of 2a) can be calculated simultaneously through GPR,
more experimental data (at least 30 as minimum number of data)
are required to accurately predict optimal conditions. Therefore,
the suitable reaction condition was estimated from temperature
and flow rate, then temperature and equivalent of 2a, followed
by flow rate and quantity of 2a.
8. (a) C. V. Calliford and K. A. Scheidt, Angew. Chem., Int. Ed., 2007,
46, 8748−8758; (b) R. Narayan, M. Potowski, Z.-J. Jia, A. P.
Antonchick and H. Waldmann, Acc. Chem. Res., 2014, 47,
1296−1310; (c) T. L. Pavlovska, R. G. Redkin, V. V. Lipson and D.
V. Atamanuk, Mol. Diversity, 2016, 20, 299–344; (d) B. Yu, D.-Q.
Yu and H. M. Liu, Eur. J. Med. Chem., 2015, 97, 673−698; (e) N. 17. This result was also supported by predicted yield from flow rate,
Ye, H. Chen, E. A. Wold, P.-Y. Shi and J. Zhou, ACS Infect. Dis.,
and quantity of 2a at 80 °C with GPy (see Supplementary Table
S5 and Fig. S5).
4 | J. Name., 2012, 00, 1-3
This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins

Page 5 of 5
ChemComm
View Article Online
DOI: 10.1039/C9CC08526B
machine learning-assisted
exploration of suitablel conditions
O
O
CO2R3
H
O
CO2R3
organocat.
flow system
+
ML
+
R1
R1
N
R2
N
minimum
R2
reactions
O
machine learning
spirooxindole analogue
A highly atom-economical enantioselective Rauhut-Currier and [3+2] annulation has been
established by flow system and machine-learning-assisted exploration of suitable conditions.