A chemometric approach to map reaction media chemoselectivity: Example of selective debenzylation

Article abstract of DOI:10.1002/ejoc.200901497

A chemometric process consisting in measuring the reactivity of a set of substrates under standardized and complementary reaction conditions was run to evaluate the possibility of building a coherent database that would give a general overview of the sele

Full text of DOI:10.1002/ejoc.200901497

SHORT COMMUNICATION
DOI: 10.1002/ejoc.200901497
A Chemometric Approach to Map Reaction Media Chemoselectivity: Example
of Selective Debenzylation
Nicolas Bensel,[a] Daniel Klär,[a] Cédric Catala,[a] Patrick Schneckenburger,[b]
Frank Hoonakker,[c] Sylvie Goncalves,[b] and Alain Wagner*[b,c]
Keywords: Chemoselectivity / Hydrogenation / Chemoinformatics / Reactivity
A chemometric process consisting in measuring the reactivity
of a set of substrates under standardized and complementary
reaction conditions was run to evaluate the possibility of
building a coherent database that would give a general over-
view of the selectivity of a variety of catalysts. This system-
atic experimental approach was applied to the hydro-
genolysis of O-benzyl ether compounds. Analysis of collected
data revealed reaction conditions with precise chemoselec-
tivity. For instance, Pd/C in EtOH or PdOH in THF enabled
the specific cleavage of the benzyl group in the presence of
p-CF3Bn groups, and addition of triethylamine in EtOH sup-
pressed the cleavage of the phenoxybenzyl bond. The latter
selectivity was exemplified on several polyfunctional sub-
strates.
Introduction
Results and Discussion
The reported approach can be seen as orthogonal to the
classical high-throughput screening (HTS) methods.[2] In-
deed, in HTS strategies, a key substrate or a key reaction is
assayed against a large number of catalysts to select the
most efficient ones. The reaction analyses usually focus on
the measurement of a single product or a single piece of
data. Accordingly, the test techniques used, such as IR ther-
mography,[3] capillary electrophoresis,[4] mass spectrome-
try,[5] sandwich immunoassay[6] and fluorescence screening
assays,[7] were optimized to achieve high throughput. On
the contrary, our approach is aimed at building a coherent
set of chemical data by assaying a set of diverse substrates
in a limited but representative number of reaction condi-
tions and collecting exhaustive data on crude mixture com-
positions.
Because of its ubiquitous use in synthesis, we decided to
use the hydrogenation reaction and more particularly the
debenzylation reaction as a model system for our study. In-
deed, the search for new selectivity in hydrogenation reac-
tions[8] and in debenzylation[9] is a continuous area of inter-
est for organic chemists. Significantly, in the course of our
bibliographic search, we noticed that many interesting pub-
lications dealing with selective hydrogenation are available
only in Japanese or Chinese language, making the use of
those results quite difficult.[10]
Achieving high levels of chemoselectivity has been the
Achilles heel of chemical synthesis.[1] The term “selectivity”
refers herein to the discrimination displayed by reagent A
when it reacts with two different reactants B and C. To ex-
trapolate possible selectivity from raw bibliographic mate-
rial it is a basic principle of mechanistic chemistry to trans-
fer structure–reactivity relationships from intermolecular
selectivity into intramolecular selectivity. However for a
chemist, the intuitive way of selecting appropriate reaction
conditions is made uneasy by the poor availability of infor-
mation about the nontransformation (stability) of func-
tional groups and also by the nonstandardization of reac-
tion media. In this paper we have investigated whether a
coherent reaction dataset generated by applying standard-
ized reaction conditions to a collection of simple substrates
would provide useful insight into reagent and media selec-
tivity and possibly the uncovering of synthetically useful
reagent selectivity.
[a] NovAliX SAS, Bioparc,
Boulevard Sébastien Brant, 67405 Illkirch-Graffenstaden,
France
[b] Laboratory of Functional ChemoSystems-UMR7199, Faculté
de Pharmacie,
74, route du Rhin, 67400 Illkirch Graffenstaden, France
Fax: +33-6-8854306
To implement the envisioned chemometric approach, our
first task was to shrink the almost infinite space of reaction
conditions to a limited number of standardized conditions.
To that end, we performed a large bibliographical analy-
E-mail: wagner@bioorga.u-strasbg.fr
[c] e-Novalys, Bioparc,
Boulevard Brant, 67400 Illkirch Graffenstaden, France
Supporting information for this article is available on the
WWW under http://dx.doi.org/10.1002/ejoc.200901497.
Eur. J. Org. Chem. 2010, 2261–2264
© 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
2261

A. Wagner et al.
SHORT COMMUNICATION
sis[11] of hydrogenation reactions and selected a set of condi- dark-green pies represent the remaining substrates. Blue
tions that sample both the most frequently used and the and red pies correspond to byproducts such as over-reduced
most diverse conditions reported in organic synthesis and compounds. Most importantly, the reliability of the set up
medicinal chemistry research. We also selected only com- was checked by random experiment reproduction.
mercially available catalysts.
We further restricted the set of catalysts, solvents and
conditions to those easily available to any bench chemists.
By applying these rules, we ended up with 18 different cata-
lysts and 8 solvents or solvent mixtures (Tables 1 and 2).
Table 1. Most representative solvents in the field of hydrogenation.
Solvents
S1
S2
S3
S4
EtOH
S5
S6
S7
S8
EtOH + 5% TFA
THF
hexane
EtOH + 5% Et3N
EtOH + 5% AcOH
EtOAc
DMF
Table 2. Most representative catalysts in the field of hydrogenation.
Catalysts
C1 Pd/C (10%)
C2 Pt/C (10%)
C3 Ru/C (10%)
C10 Pd (OH)2/C (20%)
C11 Pd/CaCO3 (5%) + Pb
C12 Pd/Al2O3 (5%)
Figure 2. Hydrogenolysis pie charts for molecules 1, 1a, 2, 2a, 3
and 3a.
C4 Pd/BaSO4 (5%)
C5 Ir/CaCO3 (5%)
C6 Rh/C (5%)
C13 Pt/C (10%) + H2O
C14 Pd(polyethyleneimine) + SiO2 (1%)
C15 Raney Ni
C7 Ni/SiO2–Al2O3 (66%)
C8 Pd (8%), Pt (2%)/C + H2O C17 Fe/graphite (5%)
C9 Cu/graphite (5%) C18 Ru/Al2O3
C16 Ir/C (1%) + H2O
The most relevant catalysts found in the literature for
hydrogenolysis are typically formed by palladium, plati-
num, nickel, iridium, iron and ruthenium metals. According
to results shown on the six master plates, palladium is by
far the most reactive metal for breaking O-benzyl ether, as
seen by the light-green colour, which is indicative of the
successful splitting of each respective benzyl-protected com-
pound (columns C1, C4, C8, C10, C12, C1a, C4a, C8a, C10a
and C12a; Figure 2). A closer look shows interesting selec-
tivity between primary benzyl 1 and primary p-trifluorome-
thylbenzyl ether 1a in ethanol (S1) with catalyst C1, C4 and
C8 or in THF (S6) and DMF (S8) with palladium hydroxide
(C10; red circles in Figure 2). Indeed, under these condi-
tions, 1 is split in high yields, whereas 1a remains mostly
without reaction. After a careful search in the literature, it
was found that such preferential hydrogenolysis was de-
scribed sketchily by the group of Spencer[12] with only very
few details. We thus set up a short experimental plan to
validate whether or not the observed selectivity would
translate to information useful with a bifunctional mole-
cule. For that, molecule 4 was synthesized and tested within
the five reaction conditions mentioned above (Table 3).
We were pleased to observe the selected reaction condi-
tions turned out to be completely selective. With both a
We hypothesized that systematically assaying limited
numbers of substrates bearing benzylic bonds with various
stereoelectronic environments would result in a data set that
would describe the reactivity and the nonreactivity of a
wider range of molecules. Our next task was thus to reduce
the chemical space by selecting a set of representative sub-
strates. As an illustration of the outcome of this approach,
a set of six molecules was chosen (Figure 1). This set of
substrates particularly focuses on primary, secondary aro-
matic O-benzyl and O-4-trifluoromethyl benzyl ether
bonds.
Figure 1. Subset of the O-benzyl ethers assayed under selected
hydrogenolysis conditions.
All possible combinations of catalysts, solvents and mo- given solvent (row S1) catalyst-induced and a given catalyst
lecules were tested under standardized conditions in a 24- (column C10) solvent-induced, selectivities could be ex-
well parallel synthesizer (MiniBlock®). Each set of 144 ex- tracted from the specifically designed data set. These first
periments carried out with substrates 1–3 and 1a–3a is pre- results appeared encouraging to us. By pushing the analysis
sented as a master plate with solvent variation in the rows of the collected data set a bit further, one can observe that
and catalyst variation in the columns (Figure 2). The com- with the corresponding secondary benzyl ether 2 and 2a,
positions of the crude reaction mixture analyzed by gas the selectivity of reaction media S1–C1, S1–C4 and S1–C8 is
chromatography are represented by using pie charts. The totally modified (highlighted by a blue circle in Figure 2).
light-green pies represent debenzylated products, whereas Whereas S1–C1 and S1–C8 reduce both ether 2 and 2a, S1–
2262
www.eurjoc.org
© 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Eur. J. Org. Chem. 2010, 2261–2264

A Chemometric Approach to Map Reaction Media Chemoselectivity
Table 3. Hydrogenolysis of bifunctional benzylated compound 4.
metries and numbers of primary, secondary alcohol and/or
phenolic functions were perbenzylated and tested with our
five most efficient catalysts (Table 4).
Catalyst
Solvent
Additive Yield[a] [%]
Pd/C, C1
EtOH, S1
THF, S6
DMF, S8
EtOH, S5
EtOH, S1
–
–
–
TFA
–
79
76
84
99
90
Pd(OH)2, C10
Pd(OH)2, C10
Pd/BaSO4, C4
Pd(8%)/Pt(2%)/C, H2O, C8
[a] Isolated yield, reaction monitored by GC and product identified
1
by H NMR spectroscopy.
C4 reduces neither of them.[13] The difficulty to predict the
selectivity switch accounts for the difficulty of extracting
knowledge from literature sources in which it is unlikely to
find a comprehensive set of substrates tested in a given cata-
lytic system and in which conditions that do not transform
a substrate are rarely reported. Further analysis revealed
clean inhibition of benzyl bond hydrogenolysis by the ad-
dition of NEt3 in the reaction media (highlighted by a red
box for rows S2). Interestingly, this nitrogen-containing
base seems to act as a poison for the catalyst for both series
1, 1a, 3 and 3a, whereas phenolic derivatives 2 and 2a do
not seem to be affected. This observation suggests the pos-
sible selective removal of the benzyl group from phenolic
substrates in the presence of alkoxybenzyls.
In all cases, only the expected product was formed in the
reaction media. However, in some reactions, the kinetics of
the reactions were found to be slower. That observation was
expected, as diffusion at the interface is greatly influenced
by the size and structure of the substrate. We circumvented
the problem by letting the reaction run for a longer length
of time. We also decided to run an experiment on a 1-g
scale in a round-bottomed flask to investigate possible up-
scale difficulties. Satisfactorily, the reaction gave a similar
yield with, however, a longer reaction time (24 h) probably
due to poor gas–liquid exchange. It thus appears that when
a catalytic system shows high chemoselectivity, transposi-
tion on complex substrates is quite straightforward. Neither
the reaction time nor the concentration seems to drastically
modify the outcome of the reaction.
To validate the synthetic opportunity of these observa-
tions we first submitted equimolar mixtures of compounds
1, 2 and 3 to the five better catalysts. As expected, only
phenolic substrate 2 was hydrogenolyzed. After that, a set
of four compounds with different molecular weights, geo-
Conclusions
Table 4. Hydrogenolysis of various benzylated compounds by the
Et3N-promoted chemoselective conditions.
To conclude, we have showed that a data set that encom-
passes reactions of simple but complementary substrates
under standardized conditions easily reveals useful chemo-
selectivity. For instance, the simple observation of specific
conditions that transformed a given substrate but did not
transform others led to the identification of robust reaction
media for selective debenzylation. Using automated plat-
forms, extension of such data collection strategies to key
organic transformations such as reduction, oxidation, C–C
coupling and protecting group chemistry would shed light
on the selectivity space that is poorly covered by existing
reagents and thus highly desirable to reach.
Entry
Catalyst
Product
Time [h] Yield[a] [%]
6a
7a
8a
9a
5
5
5
87
90
90
79
1
P/C
15
6a
7a
8a
9a
5
5
5
95
86
95
82
2
3
4
5
Pd(8%)/Pt(2%)/C
Pd(OH)2/C
Pd/Al2O3
15
6a
7a
8a
9a
5
5
5
95
91
94
76
Supporting Information (see footnote on the first page of this arti-
cle): Experimental procedure for data collection, reproducibility of
hydrogenolysis platform, procedures for the synthesis of substrates
and analytical data of new compounds; additional references.
15
6a
7a
8a
9a
15
5
15
15
83
60
63
42
6a
7a
8a
9a
15
5
15
15
79
72
72
39
Acknowledgments
Pd/BaSO4
The authors thanks the Alsace Region for financial support and
the European Comission, Marie Curie RTN-CT-2003-505020
IBAAC for financial support of D. K. We also thank Mettler-
1
[a] Isolated yield, products identified by H NMR spectroscopy.
Eur. J. Org. Chem. 2010, 2261–2264
© 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.eurjoc.org
2263

A. Wagner et al.
SHORT COMMUNICATION
Toledo for providing us a prototype of their Quantos® powder
dosing system.
[8] a) P. Tundo, A. Perosa, Reactive Funct Polym. 2003, 54, 95–
101; b) A. Mori, T. Mizusaki, M. Kawase, T. Maegawa, Y.
Monguchi, S. Takao, Y. Takagi, H. Sajiki, Adv. Synth. Catal.
2008, 350, 406–410; c) A. Mori, T. Mizusaki, Y. Miyakawa, E.
Ohashi, T. Haga, T. Maegawa, Y. Monguchi, H. Sajiki, Tetra-
hedron 2006, 62, 11925–11932.
[9] a) J. Madsen, C. Viuf, M. Bols, Chem. Eur. J. 2000, 6, 1140–
1146; b) K. R. Rawal, S. Rani, N. Kumari, Y. D. Vankar, J.
Org. Chem. 2009, 74, 5349–5355; c) L. Zhou, W. Wang, L. Zuo,
S. Yao, W. Wang, W. Duan, Tetrahedron Lett. 2008, 49, 4876–
4878.
[10] a) M. Akinori, S. Hironao, Gifu Yakka Daigaku Kiyo 2008, 57,
45–53; b) O. Keiji, S. Mutsumi, Yuki Gosei Kagaku Kyokaishi
2007, 65, 379–381.
[11] a) P. N. Rylander in Best Synthetic Methods – Hydrogenation
Methods, Academic Press, Harcourt Brace and Company Pub-
lishers, London, 1985, ch. 1–11; b) R. L. Augustine in Hetero-
geneous Catalysis for Synthetic Chemist, Dekker, New York,
1995; c) M. Frielander in Practical Catalytic Hydrogenation:
Techniques and Applications, Wiley, New York, 1971.
[1]
[2]
N. A. Afagh, A. K. Yudin, Angew. Chem. Int. Ed. 2010, 49,
262–310.
a) M. T. Reetz, Angew. Chem. Int. Ed. 2002, 41, 1335–1338; b)
J. G. De Vries, A. H. M. De Vries, Eur. J. Org. Chem. 2003,
799–811.
[3]
a) S. J. Taylor, J. P. Morken, Science 1998, 280, 267–270; b)
M. T. Reetz, M. H. Becker, K. M. Kühling, A. Holtzwarth, An-
gew. Chem. Int. Ed. 1998, 37, 2647–2650.
[4] a) M. T. Reetz, K. M. Kühling, A. Deege, H. Hinrichs, D.
Belder, Angew. Chem. Int. Ed. 2000, 39, 3891–3893; b) Y.
Zhang, X. Gong, H. Zhang, R. C. Larock, E. S. Yeung, J.
Comb. Chem. 2000, 2, 450–452.
[5] a) M. T. Reetz, M. H. Becker, H. W. Klein, D. A. Stöckigt, An-
gew. Chem. Int. Ed. 1999, 38, 1758–1761; b) J. Guo, J. Wu, G.
Siuzdak, M. G. Finn, Angew. Chem. Int. Ed. 1999, 38, 1755–
1758; c) J. W. Szewczyk, R. L. Zuckerman, R. G. Bergman, [12] M. J. Gaunt, J. Yu, J. B. Spencer, J. Org. Chem. 1998, 63, 4172–
J. A. Ellman, Angew. Chem. Int. Ed. 2001, 40, 219–219; d) C.
Market, A. Pfaltz, Angew. Chem. Int. Ed. 2004, 43, 2498–2500.
[6] P. Vicennati, N. Bensel, A. Wagner, C. Créminion, F. Taran,
Angew. Chem. Int. Ed. 2005, 44, 2–6.
[7] a) F. Badalassi, D. Wahler, G. Klein, P. Crotti, J.-L. Reymond,
Angew. Chem. Int. Ed. 2000, 39, 4067–4070; b) K. H. Shaugh-
nessy, P. Kim, J. F. Hartwig, J. Am. Chem. Soc. 1999, 121,
2123–2132; c) F. Tanaka, R. Thayumanavan, C. F. Barbas III,
J. Am. Chem. Soc. 2003, 125, 8523–8528.
4173.
[13] An equimolar mixture of substrates 2 and 2a was reacted under
the selected conditions. The results were in line with expecta-
tions according to the analysis in Figure 1. S1–C1, S1–C8 and
S8–C10 gave mostly complete reduction of both substrates. S1–
C4 gave less than 10% reduction and S6–C10 resulted in a mix-
ture of starting materials and reduced compounds.
Received: December 23, 2009
Published Online: March 10, 2010
2264
www.eurjoc.org
© 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Eur. J. Org. Chem. 2010, 2261–2264