CCLab - A multi-objective genetic algorithm based combinatorial library design software and an application for histone deacetylase inhibitor design

Article abstract of DOI:10.1016/j.bmcl.2012.05.123

The introduction of the multi-objective optimization has dramatically changed the virtual combinatorial library design, which can consider many objectives simultaneously, such as synthesis cost and drug-likeness, thus may increase positive rates of biological active compounds. Here we described a software called CCLab (Combinatorial Chemistry Laboratory) for combinatorial library design based on the multi-objective genetic algorithm. Tests of the convergence ability and the ratio to re-take the building blocks in the reference library were conducted to assess the software in silico, and then it was applied to a real case of designing a 5 × 6 HDAC inhibitor library. Sixteen compounds in the resulted library were synthesized, and the histone deactetylase (HDAC) enzymatic assays proved that 14 compounds showed inhibitory ratios more than 50% against tested 3 HDAC enzymes at concentration of 20 μg/mL, with IC₅₀ values of 3 compounds comparable to SAHA. These results demonstrated that the CCLab software could enhance the hit rates of the designed library and would be beneficial for medicinal chemists to design focused library in drug development (the software can be downloaded at: http://202.127.30.184:8080/drugdesign.html).

Full text of DOI:10.1016/j.bmcl.2012.05.123

Bioorganic & Medicinal Chemistry Letters 22 (2012) 4540–4545
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
CCLab—a multi-objective genetic algorithm based combinatorial library
design software and an application for histone deacetylase inhibitor design
Guanghua Fang a, , Mengzhu Xue a, , Mingbo Su b, Dingyu Hu a, Yanlian Li a, Bing Xiong a, , Lanping Ma a,
⇑
a,
Tao Meng a, Yuelei Chen a, Jingya Li b, Jia Li b, , Jingkang Shen
⇑
⇑
a State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, 555 Zuchongzhi Road, Zhangjiang Hi-Tech Park, Pudong,
Shanghai 201203, China
b National Drug Screening Center, 189 Guoshoujing Road, Zhangjiang Hi-Tech Park, Pudong, Shanghai 201203, China
a r t i c l e i n f o
a b s t r a c t
Article history:
The introduction of the multi-objective optimization has dramatically changed the virtual combinatorial
library design, which can consider many objectives simultaneously, such as synthesis cost and drug-like-
ness, thus may increase positive rates of biological active compounds. Here we described a software
called CCLab (Combinatorial Chemistry Laboratory) for combinatorial library design based on the
multi-objective genetic algorithm. Tests of the convergence ability and the ratio to re-take the building
blocks in the reference library were conducted to assess the software in silico, and then it was applied
to a real case of designing a 5 ꢀ 6 HDAC inhibitor library. Sixteen compounds in the resulted library were
synthesized, and the histone deactetylase (HDAC) enzymatic assays proved that 14 compounds showed
Received 20 March 2012
Revised 30 May 2012
Accepted 31 May 2012
Available online 7 June 2012
Keywords:
Combinatorial library design
Multi-objective optimization
Histone deactetylase
Inhibitor
inhibitory ratios more than 50% against tested 3 HDAC enzymes at concentration of 20 lg/mL, with IC50
values of 3 compounds comparable to SAHA. These results demonstrated that the CCLab software could
enhance the hit rates of the designed library and would be beneficial for medicinal chemists to design
focused library in drug development (the software can be downloaded at: http://202.127.30.184:8080/
drugdesign.html).
Ó 2012 Elsevier Ltd. All rights reserved.
In the past decades, the advent of combinatorial chemistry has
dramatically changed the drug discovery process, making it possi-
ble to perform a parallel synthesis a large number of chemical
compounds for bioactivity assays.1–4 However, it is not economic
to synthesize a fully enumerated library, especially with an abun-
dance of available building blocks.5–8 In the meantime, although
some leads have been identified by this approach, many of them
failed in the following pharmacokinetics evaluations.9 Hence,
virtual combinatorial library design was introduced in order to re-
duce synthesis cost and increase the hit rates by applying certain
filters or constraints during the library design process.10–13
To support the evaluation of multiple properties that medicinal
chemists were interested in, library design has evolved to apply the
multi-objective optimization technology.14 Multi-objective optimi-
zation is a strategy that considers a number of objectives simulta-
neously during the library design phase, and ultimately yields a
population of multi-dimensional solutions, each of which balances
the defined objectives. There are some privately-owned software
products to support this tactic. Among them, work from Gillet
et al. is noteworthy. Previously, they developed a genetic algorithm
based program SELECT with a fitness function by combining
several weighted objectives.15 Soon after, they reported an updated
program MoSELECT, which was based on a multi-objective genetic
algorithm and
a
fitness function based on the Pareto
algorithm.16–18 These software programs took a multi-component
sequent linking method to build the molecules, in which building
blocks would be connected step by step according to the defined
reaction sequence.
Here we described a combinatorial library design software
CCLab (Combinatorial Chemistry Laboratory), which utilizes the
multi-objective genetic algorithm based upon Pareto evaluation
combined with two fragment connection modes, namely sequent
linking and simultaneous linking. In this package, both the synthe-
sis feasibility and evaluation of ‘drug-like’ properties could be con-
sidered, and the program was organized in a more flexible pattern,
convenient to incorporate other programs to evaluate custom
properties later.
The CCLab package consists of five modules that communicate
with each other but perform different functions. As concretely
illustrated in Figure 1, the Input module parses parameters from
the input file and passes them to other modules; the CCLib module
is responsible for assembling the fragments into molecules and
building libraries on line; the CCScore module evaluates multiple
⇑
Corresponding authors. Tel.: +86 2150806600x5412; fax: +86 2150807088.
E-mail addresses: bxiong@mail.shcnc.ac.cn (B. Xiong), jli@mail.shcnc.ac.cn (J. Li),
jkshen@mail.shcnc.ac.cn (J. Shen).
These authors contributed equally to this work.
0960-894X/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.bmcl.2012.05.123

G. Fang et al. / Bioorg. Med. Chem. Lett. 22 (2012) 4540–4545
4541
In the properties calculation phase, CCScore module can con-
duct similarity calculation, diversity calculation, synthesis feasibil-
ity calculation and druglikeness calculation by calling external
programs. Specifically, all the score expressions are formatted with
the aim to minimize the parameters associated with the objectives.
In details, the library similarity is to assess the resemblance be-
tween compounds in a generated library and the reference library,
which is summarized the minimum score of each compound with
respect to any compound in reference library, then divided this
summary with compound number of the generated library to scale
the similarity in the range 0–1. Currently, the CCScore can calculate
two kinds of similarity, one is 2D fingerprint similarity and the
other is 2D pharmacophore similarity calculated with ChemAxon
program GenerateMD. The calculation of diversity of a library is
implemented to average the dissimilarity of each pairwise com-
pounds in the generated library. As for the drug-likeness calcula-
tion, a statistics-based method called Z score was adopted to
calibrate it. First, some physico-chemical properties to describe
drug-likeness were calculated for compounds in the reference
and generated libraries. Secondly, normal distribution was mod-
eled for each property k derived from the reference library and
the mean lk with the standard deviation rk was obtained. It should
Figure 1. The flow chart of the functional modules in CCLab.
be noted that the normal distribution may not fulfill all situations,
and the users can modify this easily with little python program-
ming skills. Thirdly, the Z scoreki value of each compound i in a gen-
erated library would be calculated by the following rules:
properties of the libraries, the MOGA module executes optimiza-
tion with the multi-objective genetic algorithm and the main mod-
ule integrates the modules above, performs the iterations of
optimization and collects the output information during the library
design process. The details of three important modules including
CCLib, CCScore and MOGA were described as follows.
For the construction of a library, CCLib module provides two
connection modes which are sufficient for most combinatorial
library design. As shown in Figure 2, it can be summarized as a cen-
tral scaffold with linking spots and functional groups to be linked,
but differentiates in the mode of the functional groups connected
to the scaffold simultaneously or successively. The program can
automatically select proper building blocks from a large list to
form a molecule by identifying the previously defined connection
manner and reaction types.
0
Z scorei ¼ fjp
ð1Þ
i
k
ꢁ
l j
k
k
r
k
if the property pk of this compound i is located within the range
i
i
lk
2rk, the Z scorek is 0; otherwise the Z scorek would be calcu-
i
lated according to the formula above. Lastly, the sum of Z scorek
values of all the compounds in the generated library was taken as
the property k derived penalty score of the generated library. The
synthesis feasibility of the generated library was evaluated by an
in-house program, which uses the machine learning method to cor-
relate compound’s synthesis feasibility with its properties, includ-
ing mass, logP, aromatic atom number, aromatic ring number,
asymmetric atom number, carbon ring number, fused aliphatic ring
Functional group R1
NO2
Reaction types: R1, O:ether
R2, C:aromc-aromc
O2N
*
O
N
O
or
*
NO2
R1
N
*
Manner 1
R2
.
.
.
O
.
.
.
or
or
R2
R1
+
O2N
Manner 2
N
Functional group R2
*
O
N
Scaffold
.
.
.
.
.
.
Figure 2. The ‘scaffold-functional groups’ connection modes used in the program.

4542
G. Fang et al. / Bioorg. Med. Chem. Lett. 22 (2012) 4540–4545
MOGA module is comprised by elementary components of a
standard genetic algorithm, mainly an initialization operator, a fit-
ness operator, a selection operator, a crossover operator and a
mutation operator.19,20 And the multi-objective evaluation method
Pareto ranking was adopted (details are provided in Supplemen-
tary data). 21 The fitness values of individuals are determined using
the concept of dominance where an individual is non-dominated if
a score of it in at least one of its objectives is not worse than oth-
ers’, finally leading to a Pareto surface filled with all non-domi-
nated solutions that are considered equivalent. And the niche
sharing method was applied into the fitness evaluation to distrib-
ute solutions into different clusters or niches, and the fitness of
solutions in the same niche would be penalized. In general, MOGA
would generate a group of non-dominated solutions with compro-
mised consideration of multi-objectives and clustered with a given
niche radius to decrease the probability of similar individuals.
In summary, by comparing with MoSelect, CCLab inherit the
principles of the multi-objective genetic algorithm and Pareto
ranking method from MoSELECT. While the linking modes, evalua-
tions of various properties and the synthesis feasibility were lar-
gely different from MoSELECT. Besides, our software can be
extended easily to implement other methods to calculate druglike
properties.
Figure 3. The result of fingerprint similarity sole-objective tests. The tests were
carried out with the optimal parameters and the similarity value in the best Pareto
ranking of every generation in each test were averaged.
number, fused aromatic ring number, hetero-ring number, tauto-
mer number and wiener index (details in Supplementary data).
Then the average synthesis feasibility score of each compound
was defined as the score of the generated library.
To assess the software utility, we first conducted in silico eval-
uations. All the functional groups were obtained from our previous
work,22 and the scaffold for evaluation was chosen with no target
Figure 4. The result of a two-objective tests consisting of the fingerprint similarity and the fingerprint diversity. The similarity and diversity values were collected from the
data in the best Pareto ranking of generation 1, 31, 101, 201, 401 and 501. Noting that the minimization is defined as the optimization trend of all the objective functions, the
lower left space represents a superior region.

G. Fang et al. / Bioorg. Med. Chem. Lett. 22 (2012) 4540–4545
4543
them, 27 HDAC inhibitors mostly in clinical phase were chosen
as representatives for calculating the drug-like properties to
enable the designed library having good PD/PK properties (The
ligand structures were listed in Table S1 and S2 in Supplemen-
tary data.).
R1
O
S
R2
N
HO NH
N
scaffold
In this application, 7 objectives were considered in the design of
HDAC inhibitor library, including fingerprint similarity, pharmaco-
phore diversity, synthetic feasibility, logP, mass, rotatable bond
number and polar surface area. Other MOGA parameters were set
to the optimal values as identified by parameter optimization
phase, and the size of resulted libraries was set to 5 ꢀ 6. The scaf-
fold was designed based on a class of HDAC inhibitors reported in a
patent.23 The final objective values were listed in Table S4 in Sup-
plementary data. Since the MOGA is stochastic modeling algo-
rithm, to better explore the solution space, CCLab was run 10
times to generate more virtual libraries, then a Pareto ranking
based python script was used to rank them. Finally, the library be-
longed to the best pareto ranking was illustrated in Figure 5 and
advised to synthesis. Similar to many known HDAC inhibitors,
the scaffold defined by us also has a hydroxamic acid group, which
is intended to form a chelate interaction with the Zn2+ atom in the
binding site of HDAC proteins.24–27 But the designed R1 and R2 are
very different from reported HDAC, which may introduce novelty
for HDAC inhibitor development.
H
N
*
R2
R1:
*
*
R2
N
H
NHR2
N
R2
R2
*
*
*
N
N
H
O
O
O
R2:
O
*
*
O
O
*
O
*
S
*
N
HN
Finally, 16 compounds in the library were synthesized28 limited
by the commercially available reagents and then were assessed by
inhibitory activity assays against HDAC1, 3, and 6.29 As listed in the
Table 1, the compounds can be divided into 4 groups by different
R1 fragments. Compounds 7a–7c were attributed to Group A and
19a–19d belonged to Group D, in which R1 fragments were both
aromatic, whereas 11a–11d belonged to Group B and 15a–15d
were divided to Group C, in which R1 fragments were both
aliphatic. Initial activity assays were indicated that, with drug
SAHA as the control compound, 14 compounds showed inhibitory
Figure 5. The composition of the final selected library. It is defined as the form of
scaffold—R1–R2, and the fragments are listed respectively with stars showing as the
point to be linked.
bias, on which reaction types of the two attachment sites were de-
fined manually. The reference library has the size of 10 ꢀ 10, and
each 10 building blocks of two reaction sites were randomly se-
lected from two 1000 building block pools respectively. During
the in silico test, we first optimized the MOGA parameters in the
software by using the fingerprint similarity as the solo objective
to be inspected. Then various combinations of parameters were
carried out for the optimization and only one parameter was al-
lowed to change in the process of one test run.
For each parameter, an optimal value would be picked out from
a group of parallel tests with other parameters fixed. The assess-
ment was executed by 2D fingerprint similarity sole-objective
tests. Figure 3 shows the average result of the similarity sole-objec-
tive test. Generally speaking, the convergence speed of the pro-
gram is superior to MoSELECT, and usually gets to an obvious
convergence at about the 300th generation. The ultimate finger-
print similarity could converge to about 0.05, leading to resulted li-
braries of 95% similarity. In addition, to libraries of a scaffold with
two attachment sites, 8 fragments of R1 and 7 fragments of R2 in
the reference could be found simultaneously in resulted libraries
of the last generation.
Consequently in the two-objective tests, each property would
be chosen and then combined with the fingerprint similarity. The
result of a group consisting of the fingerprint similarity and the
fingerprint diversity is illustrated in Figure 4 as an example. Com-
pared with the sole-objective test, it still shows a similar conver-
gence, resulting in that the fingerprint similarity initialized at
lib_sim 0.3 but stabilizes at 0.125 with a slight fluctuation step
by step. Similarly, 5 fragments of R1 and 6 fragments of R2 in the
reference are found simultaneously in resulted library of the last
generation (The optimized parameters were listed in Supplemen-
tary data Table S3).
ratios more than 50% at the concentration of 20 lg/mL against 3
HDAC enzymes. From IC50 values, 3 compounds were comparable
to SAHA against HDAC6. And in details, compounds showed better
activities against HDAC enzymes when they containing aromatic
rings in both R1 and R2 building blocks. It is also noted that the
compounds 11, 15, 19 and 23 all contain an amine group. By com-
paring with compounds 7, it is found that the linking part with
amide group may be better for HDAC6 interactions. And further
follow-up development of this series of HDAC inhibitors will be re-
ported elsewhere.
Multi-objective optimization has greatly changed the virtual
combinatorial library design, which can consider many properties
simultaneously. In this report, we developed a multi-objective ge-
netic algorithm based combinatorial library design software pack-
age CCLab (Combinatorial Chemistry Laboratory). The software
incorporates molecular similarity, synthesis feasibility and ‘lead-
like’ properties into the multi-objective evaluation, and uses the
genetic algorithm to implement the optimization.
In silico tests using an in house training set were carried out to
assess the software. The results indicated the software can con-
verge in the reasonable time scale about 5–10 h on an Intel XEON
2.8G processor for 10 times run of HDAC inhibitor design. From
these tests, it was found that the CCLab can find most of the pre-in-
serted fragments. Furthermore, the software was applied for de-
sign of
a HDAC inhibitor combinatorial library. Finally, 16
compounds were synthesized and evaluated by bioactivity assays.
Among them, 14 compounds showed moderate inhibitory poten-
cies against tested 3 HDAC enzymes, some were exhibited selectiv-
In addition to the in silico tests, a real application was
conducted to verify the utility of the CCLab by designing a
focused library of HDAC inhibitors. In present work, 49 reported
HDAC inhibitors were collected as the reference library. Among
ity against HDAC6, and
3 compounds have the IC50 values
comparable to the positive control marketed drug SAHA. Clearly,
the CCLab software can enhance the hit rates and would be bene-
ficial for combinatorial library design.

4544
G. Fang et al. / Bioorg. Med. Chem. Lett. 22 (2012) 4540–4545
Table 1
The activities of the 16 compounds against HDAC1, 3, and 6a
Codes
SAHA
Structure
HDAC1 IC50
0.18 0.03
(l
m)
HDAC3 IC50
0.14 0.02
(lm)
HDAC6 IC50
0.12 0.01
(lm)
O
H
N
OH
N
H
O
H
N
O
N
7a
7b
7c
9.70 1.83
22.5 6.2
9.38 1.03
16.07 3.25
2.66 0.53
0.39 0.07
0.45 0.04
0.12 0.03
N
O
S
HN OH
O
O
H
N
O
N
S
O
N
HN OH
O
O
O
S
N
O
N
2.59 0.27
S
HN OH
N
H
O
N
S
H
N
N
11a
11b
11c
NA
NA
58.67 13.39
13.46 1.73
8.77 1.55
OH
HN
O
O
N
S
H
N
N
39.43 14.25
10.32 2.18
13.48 4.14
6.62 1.37
OH
HN
O
O
O
N
S
H
N
N
OH
HN
O
O
N
S
H
N
N
11d
15a
15b
5.67 0.97
NA
5.09 0.53
NA
3.39 0.68
24.75 9.06
23.18 3.19
S
HN OH
O
O
N
O
N
S
HN OH
O
N
H
N
S
O
N
40.12 8.83
26.39 6.28
HN OH
N
H
O
N
S
O
N
HN OH
N
H
15c
35.07 6.96
34.12 5.12
9.55 0.91
O
O
N
S
O
N
15d
19a
S
6.08 1.88
4.89 1.33
2.69 0.40
HN OH
N
H
H
N
N
O
N
S
65.71 8.49
19.20 3.54
13.50 2.38
O
N
HN OH
H
N
N
O
N
S
19b
19c
6.8 1.76
6.8 0.54
4.80 0.91
5.7 0.65
1.07 0.05
1.55 0.31
O
N
HN OH
O
H
N
N
O
N
S
O
N
HN OH
H
N
N
O
N
S
S
19d
23a
2.96 0.99
1.94 0.64
2.19 0.26
1.33 0.32
1.22 0.33
O
N
N
HN OH
O
N
S
O
N
11.49 1.60
HN OH
a
IC50 values were obtained if the inhibition rate was larger than 50%. NA, not determined. Red colored parts in compounds are R1 group, while R2 were colored as blue.
from National Science & Technology Major Project ‘‘Key New Drug
Creation and Manufacturing Program’’ of China (Grant No.
2009ZX09501-010), Program of Excellent Young Scientist of
Chinese Academy of Sciences (Grant No. KSCX2-EW-Q-3-01) and
Acknowledgment
This work is supported by the State Key Laboratory of Drug
Research. The authors gratefully acknowledge financial support

G. Fang et al. / Bioorg. Med. Chem. Lett. 22 (2012) 4540–4545
4545
24. Kazantsev, A. G.; Thompson, L. M. Nat. Rev. Drug Disc. 2008, 7, 854.
25. Minucci, S.; Pelicci, P. G. Nature 2006, 6, 38.
26. Monneret, C. Eur. J. Med. Chem. 2005, 40, 1.
the National Natural Science Foundation of China (Grant 81072580
and 81102306).
27. Paris, M.; Porcelloni, M.; Binaschi, M.; Fattori, D. J. Med. Chem. 2008, 51, 1505.
28. The reagents (chemicals) were purchased from Lancaster, Acros, and Shanghai
Chemical Reagent Co. and used without further purification. Analytical thin-
Supplementary data
layer chromatography (TLC) was HSGF 254 (150–200 lm thickness; Yantai
Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.bmcl.2012.
05.123.
Huiyou Co., China). Nuclear magnetic resonance (NMR) spectroscopy was
performed on a Bruker AMX-400 and AMX-300 NMR (IS as TMS). Chemical
shifts were reported in parts per million (ppm, d) downfield from
tetramethylsilane. Proton coupling patterns were described as singlet (s),
doublet (d), triplet (t), quartet (q), multiplet (m), and broad (br). Low- and
high-resolution mass spectra (LRMS and HRMS) were given with electric,
electrospray, and matrix-assisted laser desorption ionization (EI, ESI, and
MALDI) produced by a Finnigan MAT-95, LCQ-DECA spectrometer and IonSpec
4.7 T. The purity of final compounds was assessed by the analytical HPLC
method and found to be >95%. An Agilent 1100 series HPLC with an Agilent
References and notes
1. Khosla, C. Curr. Opin. Biotechnol. 1996, 7, 219.
2. Fauchere, J.-L.; Boutin, J. A.; Henlin, J.-M.; Kucharczyk, N.; Ortuno, J.-C. Chemom.
Intell. Lab. Syst. 1998, 43, 43.
3. Hijfte, L. V.; Marciniak, G.; Froloff, N. J. Chromatogr., B: Biomed. Sci. Appl. 1999,
725, 3.
4. Williard, X.; Pop, I.; Horvath, D.; Bourel, L.; Melnyk, P.; Deprez, B.; Tartar, A. Eur.
J. Med. Chem. 1996, 31, 87.
5. Furka, A. Drug Discovery Today 2002, 7, 1.
6. Gray, N. S. Curr. Opin. Neurobiol. 2001, 11, 608.
7. Lobanov, V. S. Trends Biotechnol. 2002, 20, 86.
8. Weber, L. Drug Discovery Today 2004, 1, 261.
9. Edwards, P. J. Drug Discovery Today 2009, 14, 108.
10. Leach, A. R.; Hann, M. M. Drug Discovery Today 2000, 5, 326.
11. Weber, L. Curr. Opin. Chem. Biol. 2000, 4, 295.
12. Li, J.; Murray, C. W.; Waszkowycz, B.; Young, S. C. Drug Discovery Today 1998, 3,
105.
Zorbax Exlipse SB-C18 (25–4.6 mm,
5 lm particle sizes) reversed-phase
column was used for analytical HPLC analyses. The elution buffer was an A/B
gradient, where A = H2O and B = CH3OH. All reactions were carried out under
dry and inert condition unless otherwise stated. Experimental details for target
compounds, some intermediates and other compounds are included in the
Supplementary data.
29. The fluorogenic histone deacetylase assay was carried out as Wegener, D. et al.
reported30. Human His6-tagged and GST-fusion HDAC proteins were expressed
in insect High5 cells using a baculoviral expression system, and purified using
Ni-NTA (QIAGEN). The deacetylase activity of rhHDACs (recombinant human
HDACs) 1 and 3 was assayed with a HDAC substrate (Ac-Lys-Tyr-Lys(
AMC), and HDAC6 was assayed with another HDAC substrate(Boc-Lys(
acetyl)-AMC). The total HDAC assay volume was 25 l and all the assay
e-acetyl)-
e-
l
components were diluted in Hepes buffer (25 mM Hepes, 137 mM NaCl,
2.7 mM KCl and 4.9 mM MgCl2, pH 8.0). The reaction was carried out in black
384-well plates (OptiPlateTM-384F, PerkinElmer). In brief, the HDAC assay
13. Waldman, M.; Li, H.; Hassan, M. J. Mol. Graphics Modell. 2000, 18, 412.
14. Schneider, G. Curr. Med. Chem. 2002, 9, 2095.
15. Gillet, V. J.; Willett, P.; Bradshaw, J.; Green, DVS J. Chem. Inf. Comput. Sci. 1999,
39, 169.
16. Gillet, V. J.; Khatib, W.; Willett, P.; Fleming, P. J.; Green, DVS J. Chem. Inf.
Comput. Sci. 2002, 42, 375.
mixture contained HDAC substrate (5–50
lM, 5 ll), rhHDAC isoforms (20–
200 nM) and inhibitor (1 l). SAHA was used as the positive control for all the
l
HDACs assay. The negative controls contained neither enzyme nor inhibitor.
The incubation time for the HDAC6 assay is 3 h, and for HDAC1, 3 is 24 h at
room temperature. The reaction was quenched with the 25 ll trypsin addition
17. Gillet, V. J.; Willett, P.; Fleming, P. J.; Green, DVS J. Mol. Graphics Modell. 2002,
20, 491.
(diluted to final concentration 0.3125%). Then the plates were incubated for
30 min at room temperature to allow the fluorescence signal to develop. At
last, the fluorescence generated was monitored at wavelengths 355 nm
(excitation) and 460 nm (emission) using Envision (PerkinElmer).
Compounds were tested in 8-dose in duplicate with two or three fold serial
dilution. And the IC50 data was calculated using the software GraphPad Prism,
and chosen the equation ‘sigmoidal dose–response (variable slope)’ for curve
fitting.
18. Wright, T.; Gillet, V. J.; Green, D. V. S.; Pickett, S. D. J. Chem. Inf. Comput. Sci.
2003, 43, 381.
19. Sheridan, R. P.; SanFeliciano, S. G.; Kearsley, S. K. J. Mol. Graphics Modell. 2000,
18, 320.
20. Chen, G.; Zheng, S.; Luo, X.; Shen, J.; Zhu, W.; Liu, H.; Gui, C.; Zhang, J.; Zheng,
M.; Puah, C.; Chen, K.; Jiang, H. J. Comb. Chem. 2005, 7, 398.
21. Konak, A.; Coit, D. W.; Smith, A. E. Reliab. Eng. Syst. Saf. 2006, 91, 992.
22. Yan, B. B.; Xue, M. Z.; Xiong, B.; Liu, K.; Hu, D. Y.; Shen, J. K. Acta. Pharmacol.
Sinica 2009, 30, 251.
30. Wegener, D.; Wirsching, F.; Riester, D.; Schwienhorst, A. Chem. Biol. 2003, 10,
61.
23. Miller, T. A.; Witter, D. J.; Belvedere, S. WO2005034880, 2005.