вход по аккаунту


Charge Interactions Do the Job A Combined Statistical and Combinatorial Approach to Finding Artificial Receptors for Binding Tetrapeptides in Water.

код для вставкиСкачать
Artificial Peptide Receptors
DOI: 10.1002/anie.200501812
Charge Interactions Do the Job: A Combined
Statistical and Combinatorial Approach to
Finding Artificial Receptors for Binding
Tetrapeptides in Water**
Carsten Schmuck,* Martin Heil, Josef Scheiber, and
Knut Baumann
Intermolecular interactions involving oligopeptides or protein fragments are important supramolecular events responsible for a variety of biochemical and medicinal processes.
Therefore, artificial peptide receptors are not only interesting
as model systems for studying the principles of the underlying
supramolecular recognition event but also as starting points
for the development of sensors or drug candidates.[1] For this
purpose strong complexation of the given target peptide by
the artificial host under physiological conditions (in water) is
necessary. This is, however, quite challenging.[2, 3] On the one
hand, hydrophobic interactions, which are especially impor[*] Prof. Dr. C. Schmuck, Dr. M. Heil
Universit&t W(rzburg
Institut f(r Organische Chemie
Am Hubland, 97074 W(rzburg (Germany)
Fax: (+ 49) 931-888-4626
J. Scheiber, Dr. K. Baumann
Universit&t W(rzburg
Institut f(r Pharmazie und Lebensmittelchemie
Am Hubland, 97074 W(rzburg (Germany)
[**] Financial support from the Fonds der Chemischen Industrie and the
Deutsche Forschungsgemeinschaft (SFB 630) is gratefully
Supporting information for this article is available on the WWW
under or from the author.
tant in water, are rather nonspecific, making the design of
selective host molecules difficult.[4] On the other hand,
electrostatic interactions such as ion pairs and H bonds,
which as a result of their specificity and in the case of H bonds
also directionality[5] are quite useful for imposing selectivity in
the formation of supramolecular complexes,[6] are significantly weakened because of competitive solvation by the
polar water molecules.[7] To the best of our knowledge no
artificial oligopeptide receptor has been reported up to now
that is capable of binding its substrate in water solely based on
H bonds in combination with ion pairing. All known receptors
for larger oligopeptides require additional hydrophobic,[8]
aromatic,[1b, 9] and/or the much stronger metal–ligand interactions[10] for efficient substrate binding.
We could recently show, for example, that one-armed
tripeptide-based cationic guanidiniocarbonyl pyrrole receptors strongly bind the anionic tetrapeptide N-Ac-l-Val-l-Vall-Ile-l-Ala-OH,[2b] a model for the C terminus of the amyloid
b-peptide Ab (1–42) and even interfere with fibril formation
of the native Ab (1–42) in vitro.[11] Complex formation in this
case is, however, critically dependent on hydrophobic interactions with the alkyl side chains. Similar observations were
also made by Kilburn et al., who screened a library of
guanidinium-based flexible tweezer receptors for tripeptide
binding.[2d] In aqueous solvent, efficient binding of the dyelabeled anionic tripeptide Glu(OtBu)-Ser(OtBu)-Val was
observed (K = 8 : 104 m 1). But the deprotected substrate
was not bound at all. This indicates again that complex
formation is mainly driven by hydrophobic interactions
(probably with the bulky tBu protecting groups). Ellman
et al. described a synthetic vancomycin analogue, a rigidified
tripeptide composed of nonnatural hydrophobic amino acids
attached to vancomycin<s carboxylate binding site; this
receptor relies heavily on hydrophobic interactions for
substrate binding.[2e]
Here, in a quantitative experimental screening of a
medium-sized but focused combinatorial receptor library 1
in combination with a statistical QSAR analysis (QSAR =
quantitative structure–activity relationship), we show that the
binding of a nonhydrophobic tetrapeptide 2 in water by an
artificial receptor 1 is possible without additional hydrophobic or metal–ligand interactions. To our knowledge this is
the first example of tetrapeptide binding in water based on a
combination of H bonds and ion pairing, and our study
underlines the power of this combined experimental and
theoretical approach for the identification of supramolecular
The general design of the receptor library 1 is shown in
Figure 1. An efficient carboxylate binding site for the C
terminus of the tetrapeptide[12, 13] is attached to a linear
tripeptide unit, allowing for the formation of a hydrogenbonded b sheet with tetrapeptide substrate 2. Interactions
between the amino acid side chains both in the substrate and
the receptor further stabilize the complex and provide the
necessary substrate selectivity. Based on this receptor design
(abbreviated CBS-AA1-AA2-AA3) facile and fast solid-phase
peptide synthesis can be used. By using a combinatorial
variation of the amino acids in the tripeptide unit, structural
diversity can be introduced.[14]
2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2005, 44, 7208 –7212
Table 1: Selected association constants Kass for the complexation of the
dansylated tetrapeptide 2 (EKAA) by receptors of type 1 (CBS-AA1-AA2AA3-resin).[a]
Figure 1. Schematic representation of complex formation between the receptor library 1 and the dansylated tetrapetide substrate 2. Dansyl = (dimethylamino)naphthalene-1-sulfonyl.
We synthesized a combinatorial receptor library 1 on
Amino-TentaGel as the solid support according to a standard
Fmoc protocol using the split-mix approach[15] in combination
with the IRORI radio-frequency tagging technology.[16] In
each of the three coupling steps the same eight amino acids
were used: Lys(Boc), Tyr(tBu), Ser(tBu), Glu(OBzl), Phe,
Val, Leu, and Trp. The resulting library had 512 members.
These specific amino acids used were chosen among the
twenty proteinogenic amino acids to provide a representative
range of varying polar, charged, and hydrophobic residues
within the final receptor library. The side chains in the final
library were deprotected on-bead using HBr (25 %) in acetic
acid to provide the receptors as charged species. As a first
target for this library the polar tetrapeptide d-Glu-l-Lys-dAla-d-Ala-OH (EKAA) was chosen. This substrate is not
only challenging because of its highly flexible and polar
character but also biologically relevant as it plays a critical
role in bacterial cell wall maturation.[17] To assess the binding
properties of the 512 combinatorial receptors in aqueous
solvents, a fluorescent and water-soluble derivative 2 of this
tetrapeptide EKAA was synthesized on Wang resin as a solid
support again using a standard Fmoc protocol. Full experimental details on the synthesis of the labeled substrate are
provided in the Supporting Information.
In a first experiment a qualitative on-bead binding assay
in water (incubation of the library with a 5 mm solution of the
labeled tetrapeptide 2 in 20 mm bis-tris buffer, pH 6.0; bistris = 2,2-bis(hydroxyethyl)iminotris(hydroxymethyl)methane) indicated that receptors of the general structure 1 are
indeed capable of binding the tetrapeptide substrate 2 under
these conditions. We then determined the binding affinities of
all of the 512 receptors in the library using an on-bead
quantitative fluorescence assay in 200 mm bis-tris buffer at
pH 6.0 in combination with a high-throughput microtiterplate reader.[2b, 18] According to this assay, the binding
affinities for tetrapeptide 2 within the library vary from
Kass = 17 100 m 1 (best receptor) to Kass < 20 m 1 (worst receptor; Table 1). This amounts to a difference in activity of more
than two orders of magnitude!
To validate the binding data obtained from the solid-phase
screening, we performed complexation studies in solution for
the selected receptors 3 and 4, which were resynthesized on
Rink-amide resin using standard Fmoc protocol. (ExperiAngew. Chem. Int. Ed. 2005, 44, 7208 –7212
Kass [m 1]
17 100
15 400
15 300
[a] Obtained from quantitative screening of the library in buffered water
(experimental errors of K are estimated to be 20 %).
mental details on the syntheses can be found in the Supporting Information.) Receptor 3 (CBS-KKF) was the most
efficient in the on-bead screening (entry 1, Table 1), whereas
4 (CBS-KYK) showed a medium affinity (entry 6). UVtitration experiments[19] (Figure 2) confirmed that these two
receptors form stable complexes with the unlabeled tetrapeptide N-Ac-EKAA-OH also in solution. Furthermore, the
calculated binding constants obtained from the nonlinear
regression analysis of the binding isotherms (Kass = 15 400 m 1
for 3 and 6200 m 1 for 4) are in good agreement with the
values obtained from the quantitative on-bead screening
assay (Kass = 17 100 m 1 and 5900 m 1, respectively).
Hence, the on-bead screening results indeed represent the
situation in free solution. An inspection of the relative
binding affinities within the library (Table 1) reveals that
complex formation is strongly correlated with the positive
charge of the receptors. Those receptors containing two lysine
residues next to the guanidinium cation form the strongest
complexes with binding affinities Kass > 104 m 1 (e.g. entries 1–
3 in Table 1), whereas overall neutral or anionic receptors do
not bind the substrate efficiently. For example, the binding
2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Figure 2. Binding isotherm for the complexation of tetrapeptide N-AcEKAA-OH by receptor 3 as obtained from a UV titration experiment
corrected for absorption changes due to dilution. The dotted line
represents the curve fitting for 1:1 complexation.
affinity drops from Kass = 15 300 m 1 for CBS-KKE (entry 3)
to Kass 100 m 1 for CBS-KEE (entry 10). This correspond to
a change in affinity by two orders of magnitude for a single
amino acid substitution, an exchange that transforms an
overall dicationic receptor into a neutral one. Hydrophobic
interactions are not important for substrate binding. Receptors with just hydrophobic side chains (e.g. entry 8 or 9) have
only modest affinities (K 1000 m 1).
These qualitative trends for the binding affinities of our
flexible receptors 1 for the tetrapeptide EKAA (2) were then
quantified using a statistical QSAR analysis. To get a rough
impression of the most important structural features determining efficient binding, a simple binary structure descriptor
was used first. The following nine different properties of the
amino acids constituting the receptor were encoded in binary
fashion: acidic (= negatively charged under physiological
conditions), acyclic, aliphatic, basic (= positively charged),
charged (= either acidic or basic), hydrophobic, large,
medium, and small.[20] Figure 3 shows the regression coefficients for these nine binary variables and each of the three
amino acids within the receptor as obtained from the
statistical analysis. It can be seen that the property basicity
(= positively charged side chain), in particular of the first
amino acid next to the carboxylate binding site, has a strong
positive effect on lg Kass, while the property acidity (= negatively charged side chain) has a negative effect, and other
properties, for example, hydrophobicity have hardly any
effect at all. This clearly shows that complex formation
between receptor 1 and the tetrapeptide 2 is strongly
dependent on charge interactions.
Next, a quantitative mathematical QSAR model was used
to predict the binding affinity of receptors that were not part
of our initial library used for the screening. To improve the
quality of the mathematical model for this purpose we first
extended the initial rough set of nine descriptors, which for
example, do not allow to distinguish between basic amino
acids such as lysine or histidine, to a set of 49 selected
physicochemical, energetic, and conformational properties.[21]
This model allows an individual description of all 20
proteinogenic amino acids. This descriptor set is a subset of
the amino acid index[22] that resulted from a cluster analysis
Figure 3. Comparison of the regression coefficients (b) for the binary
descriptors. The charge of the amino acid (AA) strongly affects
complex stability. For basic AAs (positive charge under physiological
conditions) Kass is greater (positive sign for b); for acidic AAs Kass is
less. All other descriptors such as hydrophobicity have hardly any
influence. In general, the AA in the first position is more important
than the other two (increased magnitude of b). d first AA, * second
AA, ! third AA.
conducted by Tomii and Kanehisa.[21, 22] A plot of the
experimental lg Kass values versus the cross-validated predictions of lg Kass is shown in Figure 4.[23]
Figure 4. Experimental (y) vs. predicted values of lg Kass (ŷCV-1).
A quantitative inspection of this elaborated mathematical
model using a variable-selection routine[24, 25] again confirmed
that tetrapeptide binding is based solely on charge interactions.[26] The descriptors encoding hydrophobicity, aromatic
character, and van der Waals interactions hardly affect the
quality of the model.
Finally, the full model (i.e. all 3 : 49 = 147 variables) was
used to predict the binding affinities of a virtual combinatorial
library containing all possible tripeptide sequences in the
receptor that can be obtained from the common 20 proteinogenic amino acids (library size n = 8000 members). This
virtual library is 15 times larger than the initial experimental
2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2005, 44, 7208 –7212
receptor library 1 used for the screening (512 members). But
even out of the 7488 virtual receptors not yet synthesized, no
receptors with significantly improved binding properties for
tetrapeptide 2 are expected based on the QSAR analysis. The
most efficient receptor in this much larger virtual library,
CBS-KKR, is predicted to be only less than three times better
than the best hit (CBS-KKF) in our initial library (K =
50 000 m 1 vs. 17 100 m 1, respectively). Hence, the additional
investment of time and money for the synthesis and screening
of this 15-times larger virtual library based on the full set of all
20 proteinogenic amino acids instead of just the focused
library based on the preselected subset of eight amino acids
used here would not have paid off.[27]
In summary, quantitative combinatorial library screening
in combination with the statistical QSAR analysis reveals
three important points: 1) Fully flexible one-armed artificical
receptors 1 efficiently bind polar tetrapeptides such as EKAA
in water. 2) For this substrate, ion-pair formation in combination with H bonds is the main driving force for complex
formation. Hence, additional hydrophobic or aromatic interactions are not necessarily required to achieve strong peptide
binding under physiological conditions. 3) Small but carefully
designed libraries (“focused libraries”) are sufficient to
explore the features of even much larger ensembles. Hence,
the size of a combinatorial library alone is not decisive as long
as the library contains the correct range of diversity for the
given problem.[28] Currently, we are exploring the potential of
such peptide receptors as identified here for the design of
selective peptide sensors.
Received: May 25, 2005
Published online: October 18, 2005
Keywords: combinatorial chemistry · receptors ·
structure–activity relationships · supramolecular chemistry
[1] a) H.-J. BMhm, G. Klebe, Angew. Chem. 1996, 108, 2750 – 2778;
Angew. Chem. Int. Ed. Engl. 1996, 35, 2588 – 2614; b) E. A.
Meyer, R. K. Castellano, F. Diederich, Angew. Chem. 2003, 115,
1244 – 1287; Angew. Chem. Int. Ed. 2003, 42, 1210 – 1250.
[2] For selected examples of artifical peptide receptors see: a) C.
Schmuck, L. Geiger, J. Am. Chem. Soc. 2004, 126, 8898 – 8899;
b) C. Schmuck, M. Heil, ChemBioChem 2003, 4, 1232 – 1238;
c) S. Rensing, T. Schrader, Org. Lett. 2002, 4, 2161 – 2164; d) K.
Jensen, T. M. Braxmeier, M. Demarcus, J. G. Frey, J. D. Kilburn,
Chem. Eur. J. 2002, 8, 1300 – 1309; e) R. Xuo, G. Greiveldinger,
L. E. Marenus, A. Cooper, J. A. Ellman, J. Am. Chem. Soc. 1999,
121, 4898 – 4899; f) M. Sirish, H.-J. Schneider, Chem. Commun.
1999, 10, 907 – 908; g) M. Davies, M. Bonnat, F. Guillier, J. D.
Kilburn, M. Bradley, J. Org. Chem. 1998, 63, 8696 – 8703; h) A.
Hossain, H. J. Schneider, J. Am. Chem. Soc. 1998, 120, 11 208 –
11 209; i) R. Breslow, Z. Yang, R. Ching, G. Trojandt, F. Odobel,
J. Am. Chem. Soc. 1998, 120, 3536 – 3537; j) J. S. Albert, M. W.
Peczuh, A. D. Hamilton, Bioorg. Med. Chem. 1997, 5, 1455 –
[3] For review articles on artificial peptide receptors see: a) M. W.
Peczuh, A. D. Hamilton, Chem. Rev. 2000, 100, 2479 – 2494;
b) H.-J. Schneider, Adv. Supramol. Chem. 2000, 6, 185 – 216;
c) H.-J. Schneider, Angew. Chem. 1993, 105, 890 – 892; Angew.
Chem. Int. Ed. Engl. 1993, 32, 848 – 850; d) T. H. Webb, C. S.
Wilcox, Chem. Soc. Rev. 1993, 22, 383 – 395.
Angew. Chem. Int. Ed. 2005, 44, 7208 –7212
[4] For reviews on hydrophobic interactions see: a) B. Widom, P.
Bhimalapuram, K. Koga, Phys. Chem. Chem. Phys. 2003, 5,
3085 – 3093; b) L. R. Pratt, A. Pohorille, Chem. Rev. 2002, 102,
2671 – 2691; c) N. T. Southall, K. A. Dill, A. D. J. Haymet, J.
Phys. Chem. B 2002, 106, 521 – 533.
[5] J. P. Glusker, Top. Curr. Chem. 1998, 198, 1 – 56.
[6] L.-J. Prins, D. N. Reinhoudt, P. Timmerman, Angew. Chem. 2001,
113, 2446 – 2492; Angew. Chem. Int. Ed. 2001, 40, 2383 – 2426.
[7] a) M. Meot-Ner, Chem. Rev. 2005, 105, 213 – 284; b) G. A.
Jeffrey, An Introduction to Hydrogen Bonding, Oxford University Press, New York, 1997; c) J. Israelachvili, Intermolecular &
Surface Forces, 2nd ed., Academic Press, London, 1992.
[8] The importance of hydrophobic interactions for peptide-peptide
interactions within model systems (“hydrophobic cluster formation”) was for example demonstrated by Kelly in his work on bsheet nucleators: a) P. Chitnumsub, W. R. Fiori, H. A. Lashuel,
H. Diaz, J. W. Kelly, Bioorg. Med. Chem. 1999, 7, 39 – 59;
b) K. Y. Tsang, H. Diaz, N. Graciana, J. W. Kelly, J. Am. Chem.
Soc. 1994, 116, 3988 – 4005; c) J. P. Schneider, J. W. Kelly, Chem.
Rev. 1995, 95, 2169 – 2187.
[9] For reviews on aromatic interactions within peptides and in
supramolecular model systems see: a) M. L. Waters, Biopolymers 2004, 76, 435 – 445; b) M. L. Waters, Curr. Opin. Chem.
Biol. 2002, 6, 736 – 741; c) C. A. Hunter, K. R. Lawson, J.
Perkins, C. J. Urch, J. Chem. Soc. Perkin Trans. 2 2001, 651 – 669.
[10] Some recent example of metal-based peptide binding: a) A.
Buryak, K. Severin, Angew. Chem. 2004, 116, 4875 – 4878;
Angew. Chem. Int. Ed. 2004, 43, 4771 – 4774; b) H. Imai, H.
Munakata, Y. Uemori, N. Sakura, Inorg. Chem. 2004, 43, 1211 –
1213; c) A. T. Wright, E. V. Anslyn, Org. Lett. 2004, 6, 1341 –
1344; d) M. Sirish, V. Chertkov, H.-J. Schneider, Chem. Eur. J.
2002, 8, 1181 – 1188; e) H. Ogoshi, T. Mizutani, Acc. Chem. Res.
1998, 31, 81 – 89.
[11] C. Schmuck, P. Frey, M. Heil, ChemBioChem 2005, 6, 628 – 631.
[12] a) C. Schmuck, V. Bickert, Org. Lett. 2003, 5, 4579 – 4581; b) C.
Schmuck, L. Geiger, Curr. Org. Chem. 2003, 7, 1485 – 1502; c) C.
Schmuck, Chem. Eur. J. 2000, 6, 709 – 718.
[13] For comprehensive reviews of anion recognition see the following: a) K. A. Schug, W. Lindner, Chem. Rev. 2005, 105, 67 – 113;
b) M. D. Best, S. L. Tobey, E. V. Anslyn, Coord. Chem. Rev.
2003, 240, 3 – 15; c) P. A. Gale, Coord. Chem. Rev. 2003, 240,
191 – 221; d) R. J. Fitzmaurice, G. M. Kyne, D. Douheret, J. D.
Kilburn, J. Chem. Soc. Perkin Trans. 1 2002, 841 – 864; e) T. S.
Snowden, E. V. Anslyn, Curr. Opin. Chem. Biol. 1999, 3, 740 –
746; f) P. D. Beer, P. Schmitt, Curr. Opin. Chem. Biol. 1997, 1,
475 – 482; g) A. Bianchi, K. Bowman-James, E. Garcia-EspaQa,
Supramolecular Chemistry of Anions, Wiley-VCH, New York,
1997; h) F. P. Schmidtchen, M. Berger, Chem. Rev. 1997, 97,
1609 – 1646; i) C. Seel, A. GalRn, J. deMendoza, Top. Curr.
Chem. 1995, 175, 101 – 132.
[14] For review articels on the use of combinatorial receptor libraries
see: a) N. Srinivasan, J. D. Kilburn, Curr. Opin. Chem. Biol.
2004, 8, 305 – 310; b) B. Linton, A. D. Hamilton, Curr. Opin.
Chem. Biol. 1999, 3, 307 – 312; c) Y. R. DeMiguel, J. M. K.
Sanders, Curr. Opin. Chem. Biol. 1998, 2, 417 – 421; d) W. C. Still,
Acc. Chem. Res. 1996, 29, 155 – 163.
[15] K. S. Lam, M. Lebl, V. Krchnak, Chem. Rev. 1997, 97, 411 – 448.
[16] A. W. Czarnik, Curr. Opin. Chem. Biol. 1997, 1, 60 – 66.
[17] a) R. D. SSssmuth, ChemBioChem 2002, 3, 295 – 298; b) K. C.
Nicolaou, C. N. Boddy, S. BrTse, N. Wissinger, Angew. Chem.
1999, 111, 2230 – 2287; Angew. Chem. Int. Ed. 1999, 38, 2096 –
2152; c) D. H. Williams, B. Bardsley, Angew. Chem. 1999, 111,
1264 – 1286; Angew. Chem. Int. Ed. 1999, 38, 1173 – 1193; d) C. T.
Walsh, S. L. Fisher, I.-S. Park, M. Prahalad, Z. Wu, Chem. Biol.
1996, 3, 21 – 28.
[18] a) P. W. Smith, G. Chang, W. C. Still, J. Org. Chem. 1988, 53,
1587 – 1590; b) S. S. Yoon, W. C. Still, Tetrahedron 1995, 51, 567 –
2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
578; c) Y. Cheng, T. Suenaga, W. C. Still, J. Am. Chem. Soc. 1996,
118, 1813 – 1814; d) R. Arienzo, J. D. Kilburn, Tetrahedron 2002,
58, 711 – 719.
a) C. S. Wilcox in Frontiers in Supramolecular Chemistry and
Photochemistry (Eds.: H.-J. Schneider, H. DSrr), VCH, Weinheim, 1990, pp. 123 – 144; b) K. Hirose, J. Inclusion Phenom.
Macrocyclic Chem. 2001, 39, 193 – 209; c) K. A. Connors, Binding Constants, Wiley, New York, 1987; d) L. Fielding, Tetrahedron 2000, 56, 6151 – 6170.
Even for these simple binary descriptors the quality of the crossvalidated model was quite satisfactory given the limited accuracy
of the experimental screening data. 72 % of the variation in the
data could be explained (R2CV-1 = 0.72, RMSEPCV-1 = 0.87, n =
476). The figures of merit for test set prediction (average over
100 splits) were virtually identical to those of the cross-validation
(R2Test,Avg = 0.72, RMSEPTest,Avg = 0.87, q = 3) underlining the
stability of the model.
a) M. M. Gromiha, M. Oobatake, A. Sarai, Biophys. Chem. 1999,
82, 51 – 67; b) M. M. Gromiha, J. Chem. Inf. Comput. Sci. 2003,
43, 1481 – 1485.
K. Tomii, M. Kanehisa, Protein Eng. 1996, 9, 27 – 36.
Since these 49 indices were measured on different scales (e.g.
molecular weight vs. isoelectric point), the variables were
autoscaled prior to analysis (i.e. subtraction of the mean and
division by the standard deviation of the variable). The resulting
model using these 49 diverse descriptors for each amino acid
(m = 147) yielded a model quality similar to the simple binary
model (R2CV-1 = 0.71, RMSEPCV-1 = 0.88, q = 3, n = 476; R2Test,Avg =
0.73, RMSEPTest,Avg = 0.85) but now allows to dinstinguish
between all 20 proteinogenic amino acids.
a) K. Baumann, H. Albert, M. von Korff, J. Chemom. 2002, 16,
339 – 350; b) K. Baumann, M. von Korff, H. Albert, J. Chemom.
2002,16, 351 – 360.
M. Stone, J. Roy. Statist. Soc. Ser. B 1974, 36, 111 – 147.
The isoelectric point of the amino acid side chain alone is
sufficient to predict the binding affinity of the receptors. Using
only this one variable for each of the three amino acids alone
results in only a slightly decreased model quality as compared to
the full set of 49 descriptors (R2CV-1 = 0.69, RMSEPCV-1 = 0.91, q =
1, n = 476; R2Test,Avg = 0.69, RMSEPTest,Avg = 0.91).
For a critical comment on library size in the context of drugdevelopment see: R. Breinbauer, I. R. Vetter, H. Waldmann, H.
Angew. Chem. 2002, 114, 3002 – 3015; Angew. Chem. Int. Ed.
2002, 41, 2878 – 2890.
A lot of effort is nowadays directed to find this correct diversity
in designing combinatorial libraries. See for example : a) G.
Chen, S. Zheng, X. Luo, J. Shen, W. Zhu, H. Liu, C. Gui, J.
Zhang, M. Zheng, C. M. Puah, K. Chen, H. Jang, J. Comb. Chem.
2005, 7, 398 – 406; b) B. K. Shoichet, Nature 2004, 432, 862 – 865;
c) S. L. Schreiber, Chem. Eng. News 2003, 81, 51 – 61.
2005 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2005, 44, 7208 –7212
Без категории
Размер файла
154 Кб
combined, water, interactions, approach, statistics, tetrapeptide, combinatorics, findings, artificial, receptors, job, binding, charge
Пожаловаться на содержимое документа