Chiral bicyclic [2.2.2] octadiene ligands for Rh-catalysed catalytic asymmetric conjugate additions to acyclic enones: A quantitative structure-property relationship

Article abstract of DOI:10.1039/c2cc17120a

A series of chiral bicyclic [2.2.2]diene ligands gave variable ee values for Rh-catalysed asymmetric conjugate addition to an acyclic enone. The interplay between electronic and steric effects was captured in a robust predictive quantitative structure-property relationship (QSPR) model for enantioselectivity. The Royal Society of Chemistry 2012.

Full text of DOI:10.1039/c2cc17120a

View Article Online / Journal Homepage / Table of Contents for this issue
Chemical Communications
www.rsc.org/chemcomm
Volume 48 | Number 27 | 4 April 2012 | Pages 3249–3364
ISSN 1359-7345
COMMUNICATION
Andrew J. Carnell et al.
Chiral bicyclic [2.2.2] octadiene ligands for Rh-catalysed catalytic asymmetric
conjugate additions to acyclic enones: a quantitative structure–property
relationship
1359-7345(2012)48:27;1-7

Dynamic Article Links
ChemComm
View Article Online
Cite this: Chem. Commun., 2012, 48, 3279–3281
www.rsc.org/chemcomm
COMMUNICATION
Chiral bicyclic [2.2.2] octadiene ligands for Rh-catalysed catalytic
asymmetric conjugate additions to acyclic enones: a quantitative
structure–property relationshipw
Yunfei Luo,z Neil G. Berry and Andrew J. Carnell*
Received 16th November 2011, Accepted 24th December 2011
DOI: 10.1039/c2cc17120a
A series of chiral bicyclic [2.2.2]diene ligands gave variable ee
values for Rh-catalysed asymmetric conjugate addition to an
acyclic enone. The interplay between electronic and steric effects
was captured in a robust predictive quantitative structure–
property relationship (QSPR) model for enantioselectivity.
order to select the best for a given transformation. A similar
situation exists in the sulfino-olefin hybrid ligand-catalyzed
conjugate addition reactions.5
The kinetic profile of the Rh-catalyzed conjugate addition is
well established both with phosphine and diene ligands, where
the transmetallation is the rate limiting step.6 While detailed
theoretical studies have been published on asymmetric induction
by phosphine and nitrogen based ligand complexes,7,8 studies on
chiral dienes have only recently appeared. DFT calculations by
Kantchev9 for the chiral diene-Rh(I)-catalysed conjugate addition
to cyclohexanone suggest that the enantioselection step is the
carborhodation and not enone coordination and Brown10 has
shown that distortions in the transition state for carborhodation
dictate enantioselectivity. These studies strongly suggest that
transfer of chiral information from the ligand to the product is
under stereoelectronic rather than purely steric control.
Chiral diene ligands, developed independently by Hayashi,
Carreira and others have attracted a great deal of interest in
asymmetric catalysis for reactions catalysed by rhodium and
iridium.1 A range of synthetically useful reactions, most
notably asymmetric conjugate addition of aryl boronates to
enones, can be achieved in high yields and enantiomeric excess
using both C1- and C2-symmetric diene ligands.2 Previous
structural modifications of bicyclic ligands,4a–e appeared to
suggest that steric factors dominate selectivity, with an empirical
model devised to explain stereochemical outcomes based on
binding orientation of the enone with respect to the alkene
substituents on the C2-symmetric diene ligands.4b,e We recently
described a chemoenzymatic approach that gave ready and
flexible access to a new series of 1,4-disubstituted C2-symmetric
bicyclic [2.2.2]diene ligands.3 The inclusion of bridgehead
1,4-methyl groups in the ligands enabled us for the first time to
note a significant electronic effect, which improved catalytic
performance and allowed us in some cases to lower the equivalents
of aryl boronic acid required. Catalysts with electron rich ligands
gave excellent activity for both cyclic and linear enone substrates
whereas enantioselectivity for linear enones was lower. However,
use of several electron deficient ligands resulted in improved
enantioselectivity.
Intrigued by the subtle and often difficult to interpret effects
at play, we examined a series of 18 bicyclic ligands for the
asymmetric conjugate addition of phenyl boronic acid 2 to the
acyclic substrate non-3-en-2-one 1 to afford (S)-ketone-3
(Table 1). Experimentally determined enantioselectivity (ee)
was used to generate a predictive quantitative structure–property
relationship (QSPR) computational model (Fig. 1). This showed
a remarkably consistent relationship between the properties of
the ligand and the enantioselectivity outcome that appears to be
applicable to acyclic but not cyclic enone substrates.
Ligands were prepared according to previously reported
methods (details in ESI).3,4b Almost all ligands tested gave
consistently high yields, however ee’s were more variable.
Ligands 4a–h (entries 1–8), containing the 1,4-diester groups,
show an increase in ee with electron withdrawing substituents
on the aryl rings. These ligands gave lower ee’s than those
reported for Hayashi’s diene ligand 9a (entry15).4b This may
have been expected based on the larger steric requirement
of the 1,4-diester groups; indeed for the cyclic substrate
2-cyclohexenone, ligand 4a gave diminished selectivity when
compared to ligand 9a or our 1,4-dimethyl ligand 5a.3 However,
for the acyclic substrate 1, the 1,4-diester ligand 4a out-performs
the 1,4-dimethyl ligand 5a significantly (74% ee vs. 52% ee;
entries 1 and 9), suggesting that electronic factors may be more
important for acyclic than for cyclic substrates. Introduction
of 3,5-(CF3)2C6H3- aryl groups in the ligand 5f (entry 10)
gave much improved selectivity (97% ee), supporting this idea.
Du2b,c and Trost2f have recently reported the use of readily
accessible chiral linear chain dienes that exhibit promising
performance in conjugate addition reactions. However, struc-
turally similar ligands show diverse results both in enantios-
electivity and reactivity. These are difficult to predict and a
library of ligands often needs to be synthesized and tested in
Department of Chemistry, University of Liverpool, Crown Street,
Liverpool, L69 7ZD, UK. E-mail: acarnell@liv.ac.uk;
Fax: +44 151 7943587; Tel: +44 151 794 3531
w Electronic supplementary information (ESI) available: 1H-NMR,
13C-NMR, IR, HRMS, HPLC; QSPR modelling. See DOI: 10.1039/
c2cc17120a
z Current address: School of Chemistry, University of Edinburgh,
West Mains Road, Edinburgh, UK.
c
This journal is The Royal Society of Chemistry 2012
Chem. Commun., 2012, 48, 3279–3281 3279

View Article Online
Table 1 Yields and ee’s for conjugate addition to enone 1a
Here, we were surprised to find a more pronounced difference in
enantioselectivity between 1,4-dimethyl ligands 5a (52% ee) and
5f (97% ee) (entries 9 and 10) than between 1,4-diester ligands 4a
(74% ee) and 4f (86% ee) (entries 1 and 6). Clearly the
introduction of the electron withdrawing 2,5-(3,5-(CF3)2C6H3)
groups can offset the detrimental effect of the 1,4-dimethyl
groups in these ligands for the transformation of acyclic
substrates.
Interestingly the ee’s obtained with ligands 4g,h and 5h,
containing the more electron donating aryl groups, dropped,
but only to levels similar to those obtained for the 2,5-phenyl
ligands 4a and 5a respectively. For Hayashi’s 1,4-unsubsti-
tuted diene 9a (94% ee)4b a small reduction in ee occurred
when introducing the electron withdrawing 4-methoxycarbonyl
group in ligand 9i (92% ee), where we would have expected an
improved ee, and a 11% reduction in ee (below the parent 9a)
was observed for ligand 9h (83% ee) containing the electron
donating aryl groups.
Other changes at the 1,4-positions, whilst keeping the
2,5-phenyl groups constant, also resulted in differences in
enantioselectivity for the acyclic substrate but not the cyclic
substrate. For example for 2-cyclohexenone, the 1,4-diester
ligand 4a and 1,4-dimethoxymethyl ligand 7a gave very similar
ee’s of 88–89%,11 whereas for the acyclic substrate 1, ligand 4a
gave 74% ee and ligand 7a gave 57% ee. Of the three in the
subgroup 6a, 7a and 8a, the hydroxymethyl substituent in
ligand 6a gave the best enantioselectivity and all three were
better than the 1,4-methyl substituted ligand 5a. Carreira’s
ligand 10 was also tested and gave 91% ee. A direct comparison
here is difficult since the 2- and 5-positions are not aryl.
Interest has grown recently in the development of predictive
computational tools for asymmetric reactions12 that do not
require a priori knowledge of the reaction mechanism and can
be used for high throughput screening of ligands. A QSPR
model (Fig. 1) was developed to relate the experimental ee
values with calculated properties of the 18 ligands shown in
Table 1. Over 800 molecular properties (descriptors) were
calculated13,14 that were based entirely on the structure of
the ligands (see ESIw for further details). Due to the uncer-
tainty of the precise catalytic conformation of the ligands in
the selectivity-determining step, the descriptors calculated
were all independent of conformation. In our approach, only
the ground state structure of the ligand is needed, with no
requirement for inclusion of the metal or information about
the transition state. This makes the calculations quick and
considerably less demanding on computational resources than
for ab initio or DFT calculations.
Entry Ln 1,4-R
2,5-Aryl
Yield Ee of (S)-3
1
2
3
4
5
6
7
8
4a
4b
4c
4d
4e
4f
4g
4h
5a
5f
5h
6a
7a
8a
9a
9i
CO2Me
CO2Me
CO2Me
CO2Me
CO2Me
CO2Me
CO2Me
CO2Me
Me
Ph
95
92
98
95
99
95
96
74
85
73
81
76
86
71
72
52
97
49
67
57
61
94
92
83
91
3-CF3C6H4
3-CH3C6H4
4-CF3C6H4
4-MeOC6H4
3,5-(CF3)2C6H3
3,5-(CH3)2C6H3
3,4,5-(MeO)3C6H2 98
Ph
3,5-(CF3)2C6H3
9
99
99
10
11
12
13
14
15
16
17
18
Me
Me
3,4,5-(MeO)3C6H2 99
Ph
CH2OH
CH2OMe Ph
CH2F
H
H
H
H, Me
98
98
95
95
99
Ph
Ph
4-(MeOCO)C6H4
3,4,5-(MeO)3C6H2 98
i-Pr, Bn 99
9h
10
a
See ESI for reaction conditions and method for ee determinations.
The multiple linear regression QSPR model was developed
using a genetic function algorithm15 with adjusted r2 as the
objective function. The genetic algorithm was used in order to
search for a very good three-descriptor model out of the very
large (450 000) number of possible three-descriptor models
that may be developed from all the calculated properties.
We only sought models that contained a maximum of three
descriptors in order to minimize chances of an erroneous good
correlation.16 A QSPR model was found that displayed extremely
good statistical parameters. For example, for the training set the
r2adj value of 0.80 indicates that 80% of the variance is explained
by this model and the F value of 24.4 indicates a high level of
Fig. 1 QSPR model showing experimentally determined versus pre-
dicted ee’s. Numbers correspond to entries in Table 1.
c
3280 Chem. Commun., 2012, 48, 3279–3281
This journal is The Royal Society of Chemistry 2012

View Article Online
significance for the model.17 The model was validated17 and
displayed excellent statistical parameters for leave-1-out cross
validation (0.70), leave-10-out cross validation r2 (0.70) and
bootstrap r2 (0.76). The same model was found consistently,
running the genetic algorithm procedure many times. This con-
vergence gives us confidence that this is an extremely good model
with respect to this data set. As can be seen graphically in Fig. 1,
this model is statistically robust. Other statistical measures and
graphs corroborate this (supporting information).
(h) C.-G. Feng, Z.-Q. Wang, P. Tian, M.-H. Xu and G.-Q. Lin,
Chem.–Asian J., 2008, 3, 1511; (i) C. Shao, H.-J. Yu, N.-Y. Wu,
C.-Q. Feng and G.-Q. Lin, Org. Lett., 2010, 12, 3820; (j) Y. Luo
and A. J. Carnell, J. Org. Chem., 2010, 75, 2057; (k) G. Pattison,
G. Piraux and H. W. Lam, J. Am. Chem. Soc., 2010, 132, 14373;
(l) A. Saxena and H. W. Lam, Chem. Sci., 2011, 2, 2326–2331.
3 Y. Luo and A. J. Carnell, Angew. Chem. Int. Ed., 2010, 49, 2750.
4 (a) R. Shintani, K. Ueyama, I. Yamada and T. Hayashi, Org. Lett.,
2004, 6, 3425; (b) Y. Otomaru, K. Okamoto, R. Shintani and
T. Hayashi, J. Org. Chem., 2005, 70, 2503; (c) T. Gendrineau,
O. Chuzel, H. Eijsberg, J.-P. Genet and S. Darses, Angew. Chem.,
Int. Ed., 2008, 47, 7669; (d) Z.-Q. Wang, C.-G. Feng, M.-H. Xu
and G.-Q. Lin, J. Am. Chem. Soc., 2007, 129, 5336;
(e) K. Okamoto, T. Hayashi and V. H. Rawal, Chem. Commun.,
2009, 4815.
5 For examples of sulfino/alkene hybrid ligands, see: (a) Y. Wang,
X. Feng and H. Du, Org. Lett., 2011, 13, 4954; (b) X. Feng,
Y. Wang, B. Wei, J. Yang and H. Du, Org. Lett., 2011, 13, 3300;
(c) T. Thaler, L. N. Guo, A. K. Steib, M. Raducan,
K. Karaghiosoff, P. Mayer and P. Knochel, Org. Lett., 2011,
13, 3182; (d) S.-S. Jin, H. Wang and M.-H. Xu, Chem. Commun.,
2011, 47, 7230; (e) W.-Y. Qi, T.-S. Zhu and M.-H. Xu, Org. Lett.,
2011, 13, 3410; (f) G. Chen, J. Gui, L. Li and J. Liao, Angew.
Chem., Int. Ed., 2011, 50, 7681.
6 (a) T. Hayashi, M. Takahashi, Y. Takaya and M. Ogasawara,
J. Am. Chem. Soc., 2002, 124, 5052; (b) A. Kina, H. Iwaamura and
T. Hayashi, J. Am. Chem. Soc., 2006, 128, 3904; (c) A. Kina,
Y. Yasuhara, T. Nishimura, H. Iwamura and T. Hayashi,
Chem.–Asian J., 2006, 1, 707.
7 For DFT analysis on Rh-diphosphine-catalyzed conjugate
addition see: (a) T. Itoh, T. Mase, T. Nishikata, T. Iyama,
H. Tachikawa, Y. Kobayashi, Y. Yamamotod and N. Miyaurad,
Tetrahedron, 2006, 62, 9610; (b) T. Korenaga, K. Osaki,
R. Maenishi and T. Sakai, Org. Lett., 2009, 11, 2325.
Examination of the QSPR model revealed that the descrip-
tors were MATS6i (a 2D auto-correlation descriptor—Moran
autocorrelation of lag 6 weighted by ionisation potential14),
MATS3m (a 2D auto-correlation descriptor—Moran auto-
correlation of lag 3 weighted by mass14) and nCt (number of
total tertiary C(sp3)14). nCt is related to the steric properties of
the ligand whereas MATS6i and MATS3m are related to the
electronic properties of the ligands. However, precise inter-
pretation of the descriptors for ‘‘manual’’ ligand design is
challenging. Thus the design of future ligands can be performed
in silico. Virtual screening using this QSPR of an in silico
generated library of candidate ligands should identify potential
ligands that will afford high ee values.
In the asymmetric rhodium-catalysed conjugate addition to
acyclic enones there is a general trend in that more electron
deficient diene ligands give better enantioselectivity and that
electronic effects are more important than steric effects. This is
corroborated by recent DFT calculations for cyclohexanone
as the substrate.9,10 However, for the acyclic substrates in
particular the variation in ee is difficult to rationalize. We have
developed a robust QSPR model that will be employed in the
future for in silico ligand design of chiral diene lignds.
8 For DFT analysis on Pd-diphosphine-catalyzed conjugate addition
see: (a) T. Nishikata, Y. Yamamoto, I. D. Gridnev and
N. Miyaura, Organometallics, 2005, 24, 5025; (b) Y. Lan and
K. N. Houk, J. Org. Chem., 2011, 76, 4905–4909.
9 E. A. B. Kantchev, Chem. Commun., 2011, 47, 10969.
10 S. Gosiewska, J. A. Raskatov, R. Shintani, T. Hayashi and
J. M. Brown, Chem.–Eur. J., 2012, 18, 80.
Notes and references
11 Y. Luo and A. J. Carnell, unpublished results.
12 (a) N. Weill, C. R. Corbeil, J. W. De Schutter and N. Moitessier,
J. Comput. Chem., 2011, 32, 2878–2889; (b) S. E. Denmark,
N. D. Gould and L. M. Wolf, J. Org. Chem., 2011, 76, 4337;
(c) J. C. Ianni, V. Annamalai, P.-W. Phuan, M. Panda and
M. C. Kozlowski, Angew. Chem., 2006, 118, 5628.
13 Talete srl, DRAGON (Software for Molecular Descriptor Calcu-
lation) Version 6.0 - 2010 - http://www.talete.mi.it/.
14 R. Todeschini and V. Consonni, Molecular Descriptors, Wiley-VCH,
2009.
15 http://www.phakiso.com/.
1 For selected reviews and highlights, see: (a) F. Glorius, Angew.
Chem., Int. Ed., 2004, 43, 3364; (b) C. Defieber, H. Grutzmacher
and E. M. Carreira, Angew. Chem., Int. Ed., 2008, 47, 4482;
(c) R. Shintani and T. Hayashi, Aldrichimica Acta, 2009, 42, 31.
2 (a) W. C. Boyd, M. R. Crimmin, L. E. Rosebrugh,
J. M. Schomaker, R. G. Bergman and F. D. Toste, J. Am. Chem.
Soc., 2010, 132, 16365; (b) X. Hu, M. Zhuang, Z. Cao and H. Du,
Org. Lett., 2009, 11, 4744; (c) X. Hu, Z. Cao, Z. Liu, Y. Wang and
H. Du, Adv. Synth. Catal., 2010, 352, 651; (d) Z. Cao and H. Du,
Org. Lett., 2010, 12, 2602; (e) Q. Li, Z. Dong and Z.-X. Yu, Org.
Lett., 2011, 13, 1122; (f) B. M. Trost, A. C. Burns and T. Tautz,
Org. Lett., 2011, 13, 4566; (g) C.-G. Feng, Z.-Q. Wang, C. Shao,
M.-H. Xu and G.-Q. Lin, Org. Lett., 2008, 10, 4101;
16 J. G. Topliss and R. P. Edwards, J. Med. Chem., 1979, 22, 1238.
17 A. Tropsha, P. Gramatica and V. K. Gombar, QSAR Comb. Sci.,
2003, 22, 69.
c
This journal is The Royal Society of Chemistry 2012
Chem. Commun., 2012, 48, 3279–3281 3281