High-throughput production of two disulphide-bridge toxins

Article abstract of DOI:10.1039/c4cc02679a

A quick and efficient production method compatible with high-throughput screening was developed using 36 toxins belonging to four different families of two disulphide-bridge toxins. Final toxins were characterized using HPLC co-elution, CD and pharmacological studies. This journal is the Partner Organisations 2014.

Full text of DOI:10.1039/c4cc02679a

ChemComm
View Article Online
View Journal | View Issue
COMMUNICATION
High-throughput production of two disulphide-
bridge toxins†
Cite this: Chem. Commun., 2014,
50, 8408
Gregory Upert,*a Gilles Mourier,a Alexandra Pastor,a Marion Verdenaud,b Doria Alili,a
´
Denis Serventa and Nicolas Gilles*a
Received 11th April 2014,
Accepted 24th May 2014
DOI: 10.1039/c4cc02679a
www.rsc.org/chemcomm
A quick and efficient production method compatible with high-
Herein, we describe a fast and efficient production of two
throughput screening was developed using 36 toxins belonging to four disulphide bridge toxins compatible with high-throughput screening.
different families of two disulphide-bridge toxins. Final toxins were 36 toxins from 4 different families with different cysteine patterns
characterized using HPLC co-elution, CD and pharmacological studies. were used to optimize and validate our method. These toxins display a
large panel of biological activities and are mainly present in the
Venomous animals have developed an arsenal of small reticulated venoms of cone snails and arthropods. The a-conotoxins that have the
peptides for their defense and predation. Based on various cysteine pattern C1C2–C3–C4 with the cysteine connectivity 1–3, 2–4
disulphide-linked scaffolds, these toxins continuously selected and (family I, 1–15, Table 1) are potential inhibitors of the nicotinic
highly refined by the natural evolution process display a low acetylcholine receptors (nACHR). The w-conotoxins with the same
immunogenicity, and remarkable proteolytic resistance1 associated pattern but the connectivity 1–4, 2–3 (family II, 16–24, Table 1) inhibit
with a high selectivity and efficacy for a variety of important the norepinephrine transporter (NET). Both targets are involved in the
membrane receptors such as ion channels2 or G-protein coupled pain transmission mechanisms and/or in various neurodegenerative
receptors.3 Due to their short sizes and compact structures, diseases (Alzheimer, Parkinson,...).9 The toxins with the pattern
these disulphide rich peptides (DRP) show great interest for phar- C1–C2–C3–C4 and the 1–3, 2–4 connectivity identical to the apamin
maceutical applications, currently therapeutic candidates or already from the bee venom (family III, 25–30, Table 1), show a very recent
on the market.4 Toxinology of the last century was mostly focused on great interest for their activities on a Parkinson disease model.10
the identification of toxin activities in order to understand venom The sarafotoxins and their derivatives with the connectivity 1–4; 2–3
toxicity. It is estimated that only 0.01% of the animal toxins that (family IV, 30–36, Table 1) are agonists of endothelin receptors that
represent a likely natural bank of 40 million of peptides are are involved in cancer and cardiovascular diseases.11
described in dedicated databases.5 Nowadays, the advent of the
Due to the presence, in these small toxins, of post-translational
complementary transcriptomic6 and proteomic7 techniques applied modifications such as disulphide bonds, amidation or hydroxyproline
to venom glands and venoms, respectively, leads to an explosion of for example, recombinant technology is not the most adapted strategy
the number of toxin sequences. While few thousands of sequences of production while chemical synthesis is, thanks to its high versatility
were known at the beginning of the century, hundreds of thousands and flexibility. Two disulphide bridge toxins can be obtained following
will be described at the end of this decade.8 This giant database of several strategies. First, linear peptides are obtained after solid-phase
DRP sequences represents a source of potential drug candidates that synthesis and HPLC purification. Then, the disulphide bridges are
have to be synthesized in order to characterize their biological usually formed stepwise in a regioselective manner using appropriate
activities. Consequently, the high-throughput production of a large orthogonal thiol protection e.g. Trt, Acm, Mmt, STmp or StBu groups
number of reticulated peptides in sufficient amount is currently the (Fig. 1).12 This strategy is time consuming with numerous purification
major limiting step between sequencing and biological screening steps and therefore is not compatible with a high-throughput produc-
and represents nowadays the next challenge.
tion. Another method is to use selenocysteine derivatives to form
selectively a diselenide bridge instead of the natural one.13 This
method respects the disulfide’s connectivity and the biological activity
but cannot be applied with toxin sequences resulting from transcrip-
tomic and proteomic data where disulphide’s connectivity is not
known. The chosen and compatible with HT screening approach
consists in the random oxidation of the crude unprotected peptide
obtained by solid phase synthesis in a specific oxidant solution.
a
´
´
´
CEA, DSV, iBiTec-S, Service d’Ingenierie Moleculaire des Proteines (SIMOPRO),
CEA Saclay, Gif sur Yvette F-91191, France. E-mail: gregory.upert@cea.fr,
nicolas.gilles@cea.fr
b CEA, DSV, iBiTec-S, Service de Pharmacologie et d’Immunoanalyse, CEA Saclay,
Gif sur Yvette F-91191, France
† Electronic supplementary information (ESI) available. See DOI: 10.1039/
c4cc02679a
8408 | Chem. Commun., 2014, 50, 8408--8411
This journal is ©The Royal Society of Chemistry 2014

View Article Online
Communication
ChemComm
Table 1 Characteristics of the toxins used in this study. The names of the
different toxins are available in the ESI
Number
Sequence
Family/pattern
1
2
3
4
5
6
7
8
ECCNPACGRHYSCa
GRCCHPACGKNYSCa
ICCNPACGPKYSCa
GCCSTPPCAVLYCa
GCCSLPPCALSNPDYCa
FNWRCCLIPACRRNHKKFCa
RDOCCYHPTCNMSNPQICa
CCHPACGKYYSCa
9
ECCNPACGRHYSCGKa
DGRCCHPACAKHFNCa
NGRCCHPACARKYNCa
ECCHPACGKHFSCa
10
11
12
13
14
15
ZSOGCCWNPACVKNRCa
GCCSHPACNVNNPHICa
GRCCHPACGKYYSCa
16
17
18
19
20
21
22
23
24
NGVCCGYKLCHOC
VGVCCGYKLCHOC
ZGVCCGYKLCHOCa
GICCGVSFCYOC
ZTCCGYRMCVOCa
GVCCGVSFCYOC
RCCGYKMCHOC
VCCGYKLCHOC
CCHSSWCKHLC
Fig. 1 Comparison of the traditional strategy and the high-throughput
strategy developed for the synthesis of two disulphide-bridge toxins.
This method allowed us to synthesize the 36 linear toxins in
one week with excellent purities (Z75%).
25
26
27
28
29
30
CNCKAPETALCARRCQQHa
IKCNCKRHVIKPHICRKICGKNa
ALCNCNRIIIPHMCWKKCGKKa
DCPPHPVPGMHKCVCLKTC
DGCPPHPVPGMHPCMCTNTC
FPRPRICNLACRAGIGHKYPFCHCRa
During the one-step folding, three theoretical combinations
of bridges can arise. The formation of secondary misfolded
entities can be minimized by tuning the components of the oxida-
tion solution. The nature and pH of the buffer or the redox couple
used are crucial parameters. 100 mM Tris-HCl and HEPES both
containing 1 mM EDTA were used as buffers at pH 8.5 and 7.5
respectively. The chosen additives were 20% acetonitrile (ACN) and
glycerol (GOL) for solubilisation purposes, 0.5 M arginine (Arg) and
guanidinium chloride (Gdn-Cl) both used as dispersing agents.17
Glutathione reduced/oxidized (GSH/GSSG) and L-cysteine/L-cystine
(Cys/C–C) couples showing slightly different redox potentials
(respectively À0.24 V and À0.22 V in water)18 were used at different
31
32
33
34
35
36
CTCKDMTDKECLYFCHQDIIW
CTCNDMTDEECLNFCHQDVIW
CSCKDMTDKECLNFCHQDVIW
CSCKDMSDKECLNFCHQDVIW
CSCADMTDKECLYFCHQDVIW
GHACYRNCWREGNDEETCKERCa
a
C-terminal amidation and O = hydroxyproline.
This strategy needs excellent synthetic yields and an appropriate ratios (1/0.1, 1/1, 0.1/1 and 1/0.1, 0.1/1 respectively). The tempera-
folding solution with the lower formation of non-natural bridges ture used was 4 1C in order to slow down the kinetics of the
(Fig. 1).12a,14 To date, no general methods were emphasized using oxidation resulting in a more stable toxin. Each folding conditions
this strategy, the studies focusing on a small number of toxins of a were analysed using HPLC for determining the yield of folding of
single family, refolded using a very limited number of conditions.15 the correct toxin amongst the other isomers.
Therefore, we developed a general screening method including
A hierarchical ascending classification19 (HCA) that aims at
60 conditions varying buffers, additives and redox couples assembling variables between them divided the conditions into
on chosen toxins covering a wide range of physico-chemical three groups showing the same folding performances for all the
parameters (length, pI, hydrophobicity (GRAVY index)).
toxins (Fig. 2a). The first group (A, Fig. 2a and Fig. S3, ESI†)
In order to obtain two bridge toxins with high purity without contains a few numbers of conditions giving the best folding yields
the need for HPLC purification, several synthesis protocols (mean value of 66%). Acetonitrile is the mostly present additive
were applied varying amino acid equivalents, nature of the within the conditions of this group with also no additive. The redox
base (Hu¨nig’s base or NMM), reaction time or resins for having couple Cys/C–C in any proportion and the couple GSH/GSSG 1/1
the fastest and more efficient method keeping reasonable costs. lead to optimum results. The second group (B, Fig. 2a and Fig. S3,
Finally, we used a fast Fmoc strategy which consists of double ESI†) represents a large mix of conditions giving low to reasonable
short coupling steps (3 min) with 5 equiv. of Fmoc protected yields (mean value of 53%). Within this group, no specific buffers
amino acid using HCTU/NMM as an activating system on a or additives or redox couples stood out with a clear impact on the
Prelude synthesizer (Protein Technologiess).16 Every step yields. Finally, the last group (C, Fig. 2a and Fig. S3, ESI†) contains
(Fmoc deprotection, wash, capping) was optimized leading to conditions showing poor yields (average of 30%) for which the
a time of 19 min for a deprotection/coupling/capping cycle. redox couple GSH/GSSG 0.1 mM/1 mM is predominant, the buffer
This journal is ©The Royal Society of Chemistry 2014
Chem. Commun., 2014, 50, 8408--8411 | 8409

View Article Online
ChemComm
Communication
between them (Fig. 2b). The pI, GRAVY index and molecular mass
of each toxin were included in this new space and lead to the
following results: molecular mass negatively correlates with the
yields of the optimal condition (vectors in the reverse direction)
where the pI varies in the same way (vectors in the same direction)
and the GRAVY index does not affect the yield (perpendicular
vectors) (Fig. 2b). This observation allows a possible prediction of
the folding yield using the optimal condition described above.
Under the universal conditions, small peptides (M o 2000 g molÀ1
)
having a pI 4 7 will be easily folded (yields 4 80%) where the
larger peptides with lower pI will give lower but still reasonable
yields (30–80%). These yields are sufficient enough for the produc-
tion or the qualitative HT screening of these toxins.
Following the folding, no HPLC separation but only C18 SPE
was carried out for the removal of the redox couple and salts. For
determining the good pairing of the resulting peptides, three
different methods were used: (i) HPLC co-elution with a reference
when it is available or (ii) determination of the general structure
using circular dichroism analysis and/or (iii) characterization
of the affinity of the final compound with their respective targets.
16 synthetic and their respective commercially available toxins
(compounds 1–7, 16–18, 25–27 and 31–33) highlighted full over-
lapping in HPLC analysis (Fig. 3a).22 These toxins were then used
as references for the CD analyses. Within each family, the CD
spectra of the synthesized toxins were compared with that of the
reference toxins (Fig. 3b).21 In all cases, they showed the same
signature regarding their sequence similarity in particular regard-
ing the size of the loops. In addition, competitive binding assays
on nACHR were carried out for the compounds 1–3, 7, 13, 15, 28
and 30 and for NET for the compounds 16 and 18. Interestingly,
all the synthesized toxins showed activity towards their respective
Fig. 2 (a) Dendrogram obtained using HCA analysis on the different
folding conditions including three groups: A (in blue) with the best
conditions, B (in red) with conditions giving intermediate folding yields
and C (in green) for low yield conditions; (b) representation of the different
folding conditions (crosses in blue, red and green for, respectively, the
groups A, B and C with the optimal condition (blue round marker)) and
the physico-chemical parameters of the different toxins (triangles) in the
principal component space PC1/PC2.
and the additives not playing a role. In conclusion, the combination targets in agreement with affinity constants reported in the
between HEPES buffer at pH 7.5, ACN as additive and Cys/C–C literature.21 These results demonstrate that our high throughput
1/0.1 as redox couple was chosen since this particular condition production protocol is highly efficient and is compatible with HT
gives folding yields of the correct toxin higher than 70% for more screening for the determination of biological activities of toxins
than 50% of the disulphide bridge toxins studied. This condition that have been described only at the sequence level.
is advantageous in term of easy-to-use and cost. In addition,
Within this study, we developed a fast and efficient method
we applied on the different conditions a principal component for the synthesis of two disulphide bridge toxins. Our strategy
analysis20 (PCA) that aims to compress the number of variables consisting in the fast solid-phase synthesis of the linear peptide
and create a new space for a better comparison of the variables bearing or not PTMs followed by a one-step oxidation leads to
Fig. 3 (a) HPLC chromatograms of the synthetic toxin a-conotoxin GI 1 (black trace), the natural form (red trace) and their equimolar mix (blue trace);
(b) circular dichroism spectrum of the a-conotoxin GI 1 (in red) and of the a-conotoxin GIA 9 (in blue); (c) affinity determination of the toxin a-conotoxin
GIA 9 on nAchR (Torpedo).
8410 | Chem. Commun., 2014, 50, 8408--8411
This journal is ©The Royal Society of Chemistry 2014

View Article Online
Communication
ChemComm
I. Kriˇzaj, J. Proteome Res., 2012, 11, 5046; (c) M. E. Peichoto,
F. L. Tavares, M. L. Santoro and S. P. Mackessy, Comp. Biochem.
Physiol., Part D: Genomics Proteomics, 2012, 7, 361; (d) M. Madrigal,
excellent yields for the synthesis of toxins from four different
families having different cysteine patterns and pairings. The
screening method for the oxidation step, carried out on a large
number of peptides, allowed us to determine a universal condition
efficient enough for obtaining a known or unknown toxin with a
good structure in a very short time. This method that will be used
within the VENOMICS FP7 project22 is a breakthrough for reprodu-
cing synthetically what venomous animals developed over millennia
and allows a fast discovery of new ligands and potential drugs based
on the transcriptomic and proteomic studies on venoms.
Support from the European FP7 VENOMICS project is grate-
fully acknowledged.
´
˜
´
L. Sanz, M. Flores-Dıaz, M. Sasa, V. Nu´nez, A. Alape-Giron and
J. J. Calvete, J. Proteomics, 2012, 77, 280–297.
8 G. F. King, M. C. Gentz, P. Escoubas and G. M. Nicholson, Toxicon,
2008, 52, 264.
9 For recent examples see (a) R. J. Lewis, S. Dutertre, I. Vetter and
M. J. Christie, Pharmacol. Rev., 2012, 64, 259; (b) R. J. Lewis, Toxicon,
2012, 59, 524; (c) S. Dutertre and R. J. Lewis, Biochem. Pharmacol.,
2006, 72, 661.
´
10 (a) D. Alvarez-Fischer, C. Noelker, F. Vulinovic, A. Gru¨newald,
C. Chevarin, C. Klein, W. H. Oertel, E. C. Hirsch, P. P. Michel and
A. Hartmann, PLoS One, 2013, 8, e61700; (b) A. J. Kallarackal,
J. M. Simard and A. M. Bailey, Behav. Brain Res., 2013, 237, 71.
11 (a) G. Mourier, M. Hajj, F. Cordier, A. Zorba, X. Gao, T. Coskun,
A. Herbet, E. Marcon, F. Beau, M. Delepierre, F. Ducancel and
D. Servent, Biochimie, 2012, 94, 461; (b) M. A. F. Hayashi, C. Ligny-
Lemaire, Z. Wollberg, M. Wery, A. Galat, T. Ogawa, B. H. Muller,
Notes and references
¨
H. Lamthanh, Y. Doljansky, A. Bdolah, R. Stocklin and F. Ducancel,
´
1 For a recent review on disulphide rich peptides, see M. Gongora-
Peptides, 2004, 25, 1243; (c) A. Becker, E. B. Dowdleb, U. Hechler,
K. Kausera, P. Donnera and W. Schleuning, FEBS Lett., 2000,
315, 100; (d) A. Bdolah, Z. Wollberg, G. Fleminger and E. Kochva,
FEBS Lett., 1989, 256, 1.
´
Benıtez, J. Tulla-puche and F. Albericio, Chem. Rev., 2014, 114, 901.
2 For recent examples on different ionic channels, see (a) C. M. Barber,
G. K. Isbister and W. C. Hodgson, Toxicon, 2013, 66, 47; (b) Z. Chen,
F. Luo, J. Feng, W. Yang, D. Zeng, R. Zhao, Z. Cao, M. Liu, W. Li,
L. Jiang and Y. Wu, PLoS One, 2013, 8, e60201; (c) S. Diochot, A. Baron,
M. Salinas, D. Douguet, S. Scarzello, A.-S. Dabert-Gay, D. Debayle,
V. Friend, A. Alloui, M. Lazdunski and E. Lingueglia, Nature, 2012,
490, 552; (d) S. Peigneur, B. Billen, R. Derua, E. Waelkens,
S. Debaveye, L. Beress and J. Tytgat, Biochem. Pharmacol., 2011,
82, 81; (e) M. Stevens, S. Peigneur and J. Tytgat, Front. Pharmacol.,
2011, 2, 71; ( f ) E. O. Gracheva, N. T. Ingolia, Y. M. Kelly, J. F. Cordero-
Morales, G. Hollopeter, A. T. Chesler, E. E. Sanchez, J. C. Perez,
12 (a) T. M. Postma, M. Giraud and F. Albericio, Org. Lett., 2012,
´
´
´
14, 5468; (b) M. Gongora-Benıtez, J. Tulla-Puche, M. Paradıs-Bas,
O. Werbitzky, M. Giraud and F. Albericio, Biopolymers, 2011, 96, 69.
13 (a) A. Walewska, T. S. Han, M.-M. Zhang, D. Yoshikami, G. Bulaj and
K. Rolka, Eur. J. Med. Chem., 2013, 65, 144; (b) A. M. Steiner,
K. J. Woycechowsky, B. M. Olivera and G. Bulaj, Angew. Chem., Int. Ed.,
2012, 51, 5580; (c) T. S. Han, M.-M. Zhang, K. H. Gowd, A. Walewska,
D. Yoshikami, B. M. Olivera and G. Bulaj, ACS Med. Chem. Lett., 2010,
1, 140; (d) M. Muttenthaler, S. T. Nevin, A. A. Grishin, S. T. Ngo,
P. T. Choy, N. L. Daly, S.-H. Hu, C. J. Armishaw, C.-I. A. Wang,
R. J. Lewis, J. L. Martin, P. G. Noakes, D. J. Craik, D. J. Adams and
P. F. Alewood, J. Am. Chem. Soc., 2010, 132, 3514; (e) L. Moroder,
´
´
J. S. Weissman and D. Julius, Nature, 2010, 464, 1006.
3 For recent examples on different GPCR receptors, see: (a) C. Petrel,
H. G. Hocking, M. Reynaud, G. Upert, P. Favreau, D. Biass,
M. Paolini-Bertrand, S. Peigneur, J. Tytgat, N. Gilles, O. Hartley,
¨
H.-J. Musiol, M. Gotz and C. Renner, Biopolymers, 2005, 80, 85.
¨
R. Boelens, R. Stocklin and D. Servent, Biochem. Pharmacol., 2013,
14 (a) M. Reinwarth, D. Nasu, H. Kolmar and O. Avrutina, Molecules,
2012, 17, 12533; (b) L. Moroder and N. Budisa, ChemPhysChem, 2010,
85, 1663; (b) C. Fruchart-Gaillard, G. Mourier, G. Blanchet, L. Vera,
´
N. Gilles, R. Menez, E. Marcon, E. Stura and D. Servent, PLoS One,
`
11, 1181; (c) C. Boulegue, H.-J. Musiol, V. Prasad and L. Moroder,
2012, 7, e39166; (c) C. Marquer, C. Fruchart-Gaillard, G. Letellier,
Chem. Today, 2006, 24, 24; (d) L. Moroder, D. Besse, H. J. Musiol,
´
E. Marcon, G. Mourier, S. Zinn-Justin, A. Menez, D. Servent and
¨
S. Rudolph-Bohner and F. Siedler, Biopolymers, 1996, 40, 207.
¨
B. Gilquin, J. Biol. Chem., 2011, 286, 31661; (d) K. Nareoja and
15 For examples see: (a) R. Gyanda, J. Banerjee, Y.-P. Chang, A. M. Phillips,
L. Toll and C. J. Armishaw, J. Pept. Sci., 2013, 19, 16; (b) A. M. Steiner and
G. Bulaj, J. Pept. Sci., 2011, 17, 1; (c) C. J. Armishaw, N. Singh, J. L. Medina-
Franco, R. J. Clark, K. C. M. Scott, R. A. Houghten and A. A. Jensen, J. Biol.
Chem., 2010, 285, 1809; (d) B. R. Green and G. Bulaj, Protein Pept. Lett.,
2006, 13, 67; (e) E. Fuller, B. R. Green, P. Catlin, O. Buczek, J. S. Nielsen,
B. M. Olivera and G. Bulaj, FEBS J., 2005, 272, 1727; ( f ) R. DeLa Cruz,
O. Buczek, G. Bulaj and F. G. Whitby, J. Pept. Res., 2003, 61, 202.
16 C. A. Hood, G. Fuentes, H. Patel, K. Page, M. Menakuru and
J. A. E. H. Park, J. Pept. Sci., 2008, 14, 97.
17 (a) J. T. Chin, S. L. Wheeler and A. M. Klibanov, Biotechnol. Bioeng.,
1994, 44, 140; (b) M. Gonnelli and G. B. Strambini, Biophys. J., 1993,
65, 131; (c) K. Tsumoto, M. Umetsu, I. Kumagai, D. Ejima, J. S. Philo
and T. Arakawa, Biotechnol. Prog., 2004, 20, 1301; (d) Y. Hagihara,
S. Aimoto, A. L. Fink and Y. Goto, J. Mol. Biol., 1993, 231, 180.
18 P. C. Jocelyn, Eur. J. Biochem., 1967, 2, 327.
¨
¨
J. Nasman, Acta Physiol., 2011, 1; (e) C. Rouget, L. Quinton, A. Maıga,
C. Gales, G. Masuyer, C. Malosse, J. Chamot-Rooke, R. Thai,
G. Mourier, E. De Pauw, N. Gilles and D. Servent, Br.
J. Pharmacol., 2010, 161, 1361; ( f ) L. Quinton, E. Girard, A. Maiga,
M. Rekik, P. Lluel, G. Masuyer, M. Larregola, C. Marquer, J. Ciolek,
´
T. Magnin, R. Wagner, J. Molgo, R. Thai, C. Fruchart-Gaillard,
G. Mourier, J. Chamot-Rooke, A. Menez, S. Palea, D. Servent and
N. Gilles, Br. J. Pharmacol., 2010, 159, 316.
4 G. F. King, Expert Opin. Biol. Ther., 2011, 11, 1469.
5 (a) F. Jungo, L. Bougueleret, I. Xenarios and S. Poux, Toxicon, 2012,
60, 551; (b) F. Jungo and A. Bairoch, Toxicon, 2005, 45, 293.
6 For recent examples, see: (a) Y. Terrat, K. Sunagar, B. G. Fry,
T. N. W. Jackson, H. Scheib, R. Fourmy, M. Verdenaud, G. Blanchet,
A. Antunes and F. Ducancel, Toxins, 2013, 5, 1948; (b) S. Dutertre, A. Jin,
Q. Kaas, A. Jones, P. F. Alewood and R. J. Lewis, Mol. Cell. Proteomics,
2013, 12, 312; (c) M. J. Margres, K. Aronow, J. Loyacano and D. R. Rokyta,
BMC Genomics, 2013, 14, 531; (d) D. R. Rokyta, K. P. Wray and
M. J. Margres, BMC Genomics, 2013, 14, 394; (e) J. R. Prashanth,
R. J. Lewis and S. Dutertre, Toxicon, 2012, 60, 470.
´
19 S. C. Johnson, Psychometrika, 1967, 32, 241.
20 H. Abdi and L. J. Williams, Wiley Interdiscip. Rev.: Comput. Stat.,
2010, 2, 433.
21 See ESI†.
7 For recent examples, see: (a) D. L. Brinkman, A. Aziz, A. Loukas,
J. Potriquet, J. Seymour and J. Mulvenna, PLoS One, 2012, 7, e47866;
(b) A. Leonardi, D. Biass, D. Kordis, R. Stocklin, P. Favreau and
22 For more information on the VENOMICS project, see http://www.
venomics.eu.
ˇ
¨
This journal is ©The Royal Society of Chemistry 2014
Chem. Commun., 2014, 50, 8408--8411 | 8411