ZINClick: A database of 16 million novel, patentable, and readily synthesizable 1,4-disubstituted triazoles

Article abstract of DOI:10.1021/ci400529h

Since Professors Sharpless, Finn, and Kolb first introduced the concept of click reactions in 2001 as powerful tools in drug discovery, 1,4-disubstituted-1,2,3-triazoles have become important in medicinal chemistry due to the simultaneous discovery by Sharpless, Fokin, and Meldal of a perfect click 1,3-dipolar cycloaddition reaction between azides and alkynes catalyzed by copper salts. Because of their chemical features, these triazoles are proposed to be aggressive pharmacophores that participate in drug-receptor interactions while maintaining an excellent chemical and metabolic profile. Surprisingly, no virtual libraries of 1,4-disubstituted-1,2,3-triazoles have been generated for the systematic investigation of the click-chemical space. In this manuscript, a database of triazoles called ZINClick is generated from literature-reported alkynes and azides that can be synthesized within three steps from commercially available products. This combinatorial database contains over 16 million 1,4-disubstituted-1,2,3-triazoles that are easily synthesizable, new, and patentable! The structural diversity of ZINClick (http://www.symech.it/ZINClick) will be explored. ZINClick will also be compared to other available databases, and its application during the design of novel bioactive molecules containing triazole nuclei will be discussed.

Full text of DOI:10.1021/ci400529h

Subscriber access provided by UCF Libraries
Article
ZINClick: a database of 16 million novel, patentable
and readily synthesizable 1,4-disubstituted triazoles
Alberto Massarotti, Angelo Brunco, Giovanni Sorba, and Gian Cesare Tron
J. Chem. Inf. Model., Just Accepted Manuscript • DOI: 10.1021/ci400529h • Publication Date (Web): 22 Jan 2014
Downloaded from http://pubs.acs.org on January 28, 2014
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Chemical Information and Modeling is published by the American Chemical
Society. 1155 Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
ZINClick: a database of 16 million novel,
patentable and readily synthesizable 1,4ꢀ
disubstituted triazoles
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Alberto Massarotti,* Angelo Brunco, Giovanni Sorba, and Gian Cesare Tron*
Dipartimento di Scienze del Farmaco, Università degli Studi del Piemonte Orientale “A.
Avogadro”, Largo Donegani 2, 28100 Novara, Italy.
KEYWORDS: 1,2,3ꢀtriazoles, Docking, Click chemistry, Virtual library, Virtual screening
ABSTRACT. Since Professors Sharpless, Finn, and Kolb first introduced the concept of the
“click reactions” in 2001 as powerful tools in drug discovery, 1,4ꢀdisubstitutedꢀ1,2,3ꢀtriazoles
have become important in medicinal chemistry due to the simultaneous discovery by
Sharpless, Fokin and Meldal of a perfect click 1,3ꢀdipolar cycloaddition reaction between
azides and alkynes catalyzed by copper salts. Due to their chemical features, these triazoles
are proposed to be aggressive pharmacophores that participate in drugꢀreceptor interactions
while maintaining an excellent chemical and metabolic profile. Surprisingly, no virtual
libraries of 1,4ꢀdisubstitutedꢀ1,2,3ꢀtriazoles have been generated for the systematic
investigation of the clickꢀchemical space. In this manuscript, a database of triazoles called
ZINClick is generated from literatureꢀreported alkynes and azides that can be synthesized
within three steps from commercially available products. This combinatorial database
contains over 16 million 1,4ꢀdisubstitutedꢀ1,2,3ꢀtriazoles that are easily synthesizable, new
ACS Paragon Plus Environment
1

Journal of Chemical Information and Modeling
Page 2 of 37
1
2
3
4
5
6
7
8
and patentable! The structural diversity of ZINClick (http://www.symech.it/ZINClick) will
be explored; ZINClick will also be compared to other available databases, and its application
during the design of novel bioactive molecules containing triazole nuclei will be discussed.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1. INTRODUCTION
Virtual screening1 is an important tool during the search for novel bioactive compounds
despite its recognized pitfalls.2 This methodology has become important within numerous
drug discovery programs in academia and industry. The opportunity to screen databases
containing over a million compounds in a relatively short period is an added bonus for drug
hunters and must be exploited to the fullest. Several review articles3 and books4 reported
successful cases where virtual screening campaigns have unveiled novel hit compounds as
starting points for classical lead optimization procedures. To date, the number of virtual
libraries, both public and private,5 available for computational chemists is growing
exponentially, increasing the regions of chemical space that can be used against a specific
biological target. Several libraries are already available for docking with 3D structures and
include the correct protonation, tautomeric and conformational states. One of the most
important databases used worldwide is ZINC. This free database includes approximately 20
million commercially available compounds; it was developed by Stoichet6 and improved by
Irwin.7 Another wellꢀknown library is the ACD (Available Chemical Directory) from
Molecular Design Limit8; although this library contains over 2.3 million compounds, it is not
public, and it is expensive.
To expand the chemical space available for docking, other authors created novel
combinatorial databases. For example, TIN is a combinatorial collection of compounds
generated by employing a definite set of multicomponent reactions.9 Our experience with
multicomponent reactions10 suggests that not all of the compounds present in that database
ACS Paragon Plus Environment
2

Page 3 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
are easily available due to the strong dependence upon multicomponentꢀderived products.
Another published database enumerates the entire virtual chemical universe of molecules
containing up to 11, 13 and 17 atoms formed by combining the following four elements:
carbon, nitrogen, oxygen, and fluorine.11 Although the types of molecules generated using
this method are interesting, usually these compounds lack readily apparent synthetic routes.
Two papers have demonstrated that medicinal chemists tend to use the same reliable and
robust reactions to build their molecular scaffolds during drug discovery.12
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Databases containing natural products have also been reported.13 In this case, the retrieval
of several natural compounds is difficult, limiting the usefulness of these databases.
Apart from the cost of the compounds, other problems may arise from the vendors.
Sometimes, the desired molecules are no longer available or their synthesis has never been
reported, rendering the use of these molecules impractical. In other cases, the amount of a
compound available is only enough for preliminary screening; some vendors have also had
reꢀsupply issues. This drawback has been highlighted by at least one paper.14 The
patentability of the compounds is another issue. Many compounds are not new, generating
problems with intellectual property. Finally, the same compounds can be ordered by
competitors working on the same projects.
These considerations show one of the most important problems that computational
chemists face when involved in virtual screening research: the rapid access to the selected
compounds to test their hypotheses. Because most computational chemists are not synthetic
chemists, they must ask to an organic chemist to collaborate if the predicted compounds are
not available. These undertakings are not easy, especially when the required synthesis is long
and challenging; synthetic chemists are notoriously reluctant to undertake long syntheses for
compounds predicted via computer.
ACS Paragon Plus Environment
3

Journal of Chemical Information and Modeling
Page 4 of 37
1
2
3
4
5
6
7
8
9
The synthetic feasibility of the selected compounds is critical during the drug discovery
process. Only a few reactions are commonly used by medicinal chemists; the most recurrent
functional group in drugs is the amide due to its accessibility.12b Finally, the number of
synthetic steps required to synthetize the molecules must be considered. With few exceptions,
drug candidates should not require more than eight synthetic steps.12a
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In 2001, Professors Sharpless, Finn, and Kolb introduced “click reactions”15 as powerful
tools for drug discovery. Afterward, 1,4ꢀdisubstitutedꢀ1,2,3ꢀtriazoles became important in
synthetic chemistry thanks to the simultaneous discovery by Sharpless, Fokin, and Meldal of
the perfect click 1,3ꢀdipolar cycloaddition reaction between azides and alkynes catalyzed by
copper salts.16 Due to their electronic nature, these triazoles might be aggressive
pharmacophores17 that can actively participate in drugꢀreceptor interactions while
maintaining an excellent chemical and metabolic profile.
New technologies require time to mature and impact research, usually between 10ꢀ15
years;18 despite the interesting features of click chemistry, a 2008 paper revealed that only
14% of the published works on click chemistry were related to drug discovery,19 indicating
the reluctance of medicinal chemists to use this reaction despite the successful results.20 We
have considered using a library of 1,4ꢀdisubstituted triazoles for virtual screening since 2009.
After a literature search, we noticed that no systematic attempt has been made to generate a
feasible click library, and no virtual libraries of 1,4ꢀdisubstitutedꢀ1,2,3ꢀtriazoles existed to
investigate the clickꢀchemical space systematically. While conducting internet research, we
found different enumerating software programs (e.g., Chemaxon reactor,21 Maestro,22
MOE,23 ...) able to generate virtual libraries of molecules that could be constructed using
click reactions. Another website provided a freely downloadable a program that generates a
library of 1,4ꢀdisubstituted triazoles after the insertion of azides and alkynes;24 however, the
overall combination output is limited to 10,000 triazoles, limiting its usefulness. In these
ACS Paragon Plus Environment
4

Page 5 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
programs, the operator must insert the desired starting materials, and the selection process for
the azides and alkynes is too arbitrary. The same concept has been published, generating
AutoClickChem.25 Even in this case, a series of reactions is listed following the click criteria,
and a library will be generated after inserting the preferred building blocks. These methods
are useful, but they are limited by the number of building blocks and the process of selecting
which building block to add.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In this manuscript, we describe the preparation of a novel library containing over 16
million 1,4ꢀdisubstitutedꢀ1,2,3ꢀtriazoles that can be synthesized using existing alkynes and
azides within three synthetic steps starting from commercially available products while using
an established synthetic procedure. Due its public nature and the inspiration provided by the
work of Stoichet6, we called this library ZINClick. Importantly, most of the molecules in this
database are new and patentable. Because they are derived from known azides and alkynes,
after identification using virtual screening and docking procedures, these compounds should
be easily prepared, enabling tests of the docking hypotheses. Using azides and alkynes
reported in the literature synthesized for a particular scope generates a library of triazoles as
diverse as possible. Many medicinal chemists tend to create molecules that are feasible
following their chemical knowledge and experience.
2. EXPERIMENTAL PROCEDURES
All molecular modeling studies were performed on a Tesla workstation equipped with two
Intel Xenon X5650 2.67 GHz processors and Ubuntu 10.04 (www.ubuntu.com). A standard
LAMP server was prepared: the website was programmed using PHP with the JSDraw
structure editor (www.scilligence.com), and a MySQL database (www.mysql.com) was used
to store the chemical structures. Different software was used to elaborate the chemical
information: PerlMol and OpenEye (www.eyesopen.com) applications. The graphs were
ACS Paragon Plus Environment
5

Journal of Chemical Information and Modeling
Page 6 of 37
1
2
3
4
5
generated using R.26 The protein structures and 3D chemical structures were generated in
PyMOL.27
6
7
8
9
2.1. Compounds source. The collection of azides and alkynes was retrieved using
SciFinder®.28 This searchable database includes chemical structures available from chemical
vendors, as well as published literature and patents. The structures were downloaded using
the webꢀbased interface and manually added to the database. To assess whether the selected
triazoles could be synthesized in a reasonable number of steps, we decided to use azides and
alkynes with reported syntheses shorter than three steps when beginning from commercially
available products. Consequently, the maximum number of synthetic steps required to
synthesize the desired compounds could not exceed seven.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Additionally, we made 500 Daltons the upper limit for the molecular weights for the azides
and the alkynes, facilitating compatibility with synthetic drugs. Therefore, the maximum
molecular weight for the final compounds cannot exceed 1000 Daltons. For chiral molecules,
if the absolute stereochemistry was not available, both enantiomers were generated. Using
these criteria, 4627 alkynes and 3506 azides were retrieved (the database was updated
December 31, 2012). The azides and alkynes were saved as canonical SMILES and stored in
a MySQL database.
2.2. Generation of ZINClick. An in house Perl script based on PerlMol29 was used to
retrieve each azide and perform a virtual combinatorial reaction with all alkynes. The
resulting SMILES for 1,4ꢀtriazoles were saved in a MySQL database. Each structure was
rebuilt as a 3D structure using OMEGA230. The molecular descriptors (listed in Supporting
Information Table S1) were calculated for each output. These calculations were performed
with the FILTER31 software from OpenEye. Finally, the database was exported in SMILES
format. For internal use, the resulting SMILES strings were stored in a MySQL database to
ACS Paragon Plus Environment
6

Page 7 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
facilitate standard Structured Query Language (SQL) searches and to prepare for future webꢀ
server integration.
7
8
9
2.3. Data curation for ZINClick. To prevent automation problems during the creation of
ZINClick, the generation steps were carefully checked. The collections of alkynes and azides
were checked for duplicate structures. For each generation process, a log file detailing any
errors was produced. All errors were corrected manually; otherwise, the compound was
deleted.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2.4. Reference databases. Three different compound databases were considered: the ZINC
“All Purchasable” subset (19,607,982 compounds dated January 11, 2013), the ZINC
“Natural product” subset (189,355 compounds) and the DrugBank “Approved Drugs” set
(1,486 compounds).32 The molecular descriptors were subsequently calculated for all
databases using FILTER. The descriptors were derived from the Lipinski “drugꢀlike”,33 the
Oprea “leadꢀlike”,34 and the ruleꢀofꢀthree “fragmentꢀlike” rules:35 molecular weight (MW),
logP(o/w), the number of hydrogen bond acceptors (HBAs), the number of hydrogen bond
donors (HBDs), the total polar surface area (TPSA) and the amounts of rotatable bonds
(rotB), charges and chiral atoms.
2.5. Properties subsets. Three different collections of ZINClick compounds were prepared
using the following criteria. (I) A drugꢀlike subset (DL) was based on Lipinski’s rules: MW
<= 500 and MW >= 150, logP <= 5, rotB <=7, PSAa < 150, HBD <= 5 and HBA <= 10.33
(II) A leadꢀlike subset (LL) was used: MW <= 350 and MW >= 250, logP <= 3.5, rotB <=
7.34 (III) A fragmentꢀlike subset (FL) was also used: logP<=3.5, MW <=250 and rotB <= 5.35
The subset collections include 4,140,042, 744,235 and 22,104 compounds, respectively. The
subsets are exported in both SMILES and SDF formats.
ACS Paragon Plus Environment
7

Journal of Chemical Information and Modeling
Page 8 of 37
1
2
3
4
5
6
7
8
2.6. 10% subset. To reduce the number of ZINClick compounds, a random selection
(10%) of the compounds in ZINClick was generated. The 10%ꢀsubset (10S) consists of
1,622,226 compounds and is available in both SMILES and SDF formats.
9
2.7. Diversity subset. To assess the structural diversity of ZINClick, a diversityꢀsubset
(DS) was prepared. First, the ChemFP Substruct fingerprints were generated for the
compounds using chemfp v1.136 with the OEChem37 and GraphSim38 Python Toolkits. This
fingerprint comprises 881 bits; they are not the same but were "heavily inspired" by the
CACTVS 661ꢀbit substructure fingerprints used in PubChem.39 Finally, the SUBSET 1.0
algorithm40 was used to select compounds that differ from those previously selected by at
least a Tanimoto cutoff of 0.7; 2,323 compounds in SMILES format were selected. To
provide a representative group of molecules for comparison with the ZINClick Diversityꢀset,
27,442 structures were selected from ZINC using the same procedure. A short perl program
was written to record the occurrence of each fingerprint for plotting. The PCA plots were
generated using R.26 The diversity subset was exported in both SMILES and SDF formats.
2.8. PAINS Analysis. PAINS filters present in the literature in SLN format (Tables S6, S7,
and S9 in the Supporting Information from the work of Baell and Holloway)41 were
considered. The ZINClick database was mapped to individual PAINS substructure motifs
using inꢀhouse perl scripts containing the SMARTS form version of the filters published by
the Guha group.42
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
3. RESULTS AND DISCUSSION
The ZINClick database provides 16,222,262 virtual 1,4ꢀdisubstituted 1,2,3ꢀtriazoles
generated using existing alkynes and azides that require no more than three synthetic steps
from commercially available products and have a molecular weight cutꢀoff of 500 Daltons.43
After the database was generated, common descriptive properties were computed to define
ACS Paragon Plus Environment
8

Page 9 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
the drugꢀlike and leadꢀlike compounds. The Lipinski33 violations for the ZINClick
compounds are reported in Figure 1. Of the 16 million compounds, 4,140,042 (25%) are
drugꢀlike, and 744,235 are leadꢀlike (5%) (the detailed count is reported in Supporting
Information Table S2). To understand these results, we computed these data using three other
databases: the ZINC “All Purchasable” subset, the ZINC “Natural product” subset and
DrugBank, representing the known chemical space, the natural product chemical space and
the drug chemical space, respectively. Clearly, the drugꢀlike character of our library is poor
compared the above libraries; in those libraries, the amounts of drugꢀlike compounds account
for, respectively, approximately 61%, 56% and 57% of the total. This result includes only
compounds that do not violate Lipinski’s rules. In our case, the choice to retrieve azides and
alkynes with a molecular weight ≤500 Da, dramatically reduced the percentage of triazoles
with a drugꢀlike molecular weight.44 If the molecules with no more than one violation are
included, about half of the ZINClick library is drugꢀlike (52%, 8,522,007 compounds).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. Histogram of the Lipinski violations as percentages of the ZINClick (orange),
ZINC (cyan), ZINC “Natural product” (green) and DrugBank (yellow) databases. The details
are reported in Supporting Information Table S2.
ACS Paragon Plus Environment
9

Journal of Chemical Information and Modeling
Page 10 of 37
1
2
3
4
5
6
7
8
Lipinski’s rules are depicted in Figure 2. After analysis, the molecular weight (MW)
distributions show a peak value between 450 and 500 Da (Figure 2a) on a curve similar to a
Gaussian distribution. The distribution of the log P values also show a Gaussian shaped curve
with a peak centered at 3.5 log P units (Figure 2b). The peaks for the hydrogen bond
acceptors (HBA) and hydrogen bond donors (HBD) are at 8 and 1, respectively, and both
curves fall off rapidly at maxima of 20 and 12, respectively (Figure 2cꢀd). The polar surface
area (PSA) and the rotatable bonds count (rotB) distribution have peaks at approximately 100
and 7, respectively (Figure 2eꢀf). Finally, approximately 68% of the compounds are neutral,
and 7,062,641 structures (43%) contain at least one stereogenic center (Figure 2gꢀh).
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment
10

Page 11 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 2. Distribution histograms for the descriptors (MW, logP, HBA, HBD, TPSA, rotB,
charges and chiral, respectively) of the compounds in ZINClick.
ACS Paragon Plus Environment
11

Journal of Chemical Information and Modeling
Page 12 of 37
1
2
3
4
5
6
7
8
The distributions for each descriptor of compounds from ZINClick and the other databases
are compared using a line graph (Figure 3). The molecular weight space covered by ZINClick
is shifted toward higher values (Figure 3a) due to the MW cutꢀoff used while preparing the
database. The distributions of the logP values are more similar (Figure 3b). The clusters are
equally populated between 2 and 5, while ZINC shows a preference for the region between 2
and 4. The number of hydrogen bond acceptors in ZINClick compounds is generally double
that of compounds from the other databases (Figure 3c). However, for hydrogen bond
donation capabilities, similar behavior was observed between all of the databases (Figure 3d).
Consequently, for the higher MW and HBD counts in ZINClick, higher polar surface area
values are generated relative to ZINC and the other databases (Figure 3e). Unsurprisingly, the
ZINClick database is very different from the Natural products and DrugBank databases; for
the latter, fewer rotatable bonds are preferred (Figure 3f). Finally, a similar netꢀcharge
distribution (Figure 3g) is apparent, while chiral molecules are more common in the Natural
products database (Figure 3h).
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment
12

Page 13 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3. Comparison of the distributed properties for the ZINClick (orange), ZINC (cyan),
ZINC “Natural product” (green) and DrugBank (yellow). The line Graphs represent the
ACS Paragon Plus Environment
13

Journal of Chemical Information and Modeling
Page 14 of 37
1
2
3
4
5
6
percentages of MW, logP, HBA, HBD, TPSA, rotB, charges and chiral molecules,
respectively, in the database.
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The diversity of the ZINClick database was analyzed versus ZINC. Molecular similarity is
a key prerequisite when assessing molecular diversity.45 Many different techniques can be
used to measure whether two compounds are similar;46 fingerprint methods are simple and
commonly applied. The chemfp36 software was used to generate a fingerprint from the
GraphSim Python Toolkits38 of OpenEye. The SUBSET algorithm40 selected 2,323
compounds using the calculated fingerprints. This diversityꢀsubset contains representative
compounds that possess a Tanimoto index below 0.7 with any selected and nonꢀselected
compounds in the database. When using this representative subset, a principal component
analysis (PCA) of the chemical space represented by ZINClick was performed. The first two
principal components accounted for 92.1% and 7.4% of the Xꢀvariance. Figure 4a shows that
all of the ZINClick compounds are distributed across the entire chemical space. The
projection of ZINC compounds in the PCA plot (see Supporting Information Figure S4)
demonstrates that the two databases do not completely overlap, indicating that ZINClick
explores new regions of chemical space. Figure 4a shows the four distinctive regions related
to the principal ZINClick subset.
To explore the structural diversity in ZINClick, the diversityꢀsubset selection was applied
to the ZINC database, generating a ZINCꢀdiversityꢀsubset containing 27,442 compounds. The
structural fingerprints used during the selection process comprised 881 bits, each representing
a discrete substructure. A table reporting how many molecules from ZINClick and ZINC
contained the substructures is included in Supporting Information Table S4. Of the 881
fingerprint substructure features, 737 appear in at least one ZINC molecule, and 649 appear
in at least one ZINClick compound. In addition, 633 of these features are common to at least
ACS Paragon Plus Environment
14

Page 15 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
one molecule from ZINC and ZINClick. Although ZINClick contains two substructures that
do not appear in ZINC, ZINC contains 81 substructures that do not appear in ZINClick. The
distribution of the fingerprints is shown in Figure 4b. Each point in the figure represents a
single substructural feature. Clearly ZINClick shares several substructure fingerprints with
ZINC at a comparable frequency.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4. Diversity of ZINClick versus ZINC. (a) A principal component analysis (PCA) plot
comparing the chemical space defined by the ZINClick databases: all compounds (orange),
the “drugꢀlike” subset (blue), the “leadꢀlike” subset (yellow) and the “fragmentꢀlike” subset
(violet). (b) Diversity of ZINClick versus ZINC. The substructural fingerprints were analyzed
in each compound from both the ZINC and ZINClick diversityꢀsubsets using an 881ꢀbit
search. Each point in this logarithmic scatter plot corresponds to the number of compounds
found with least one copy of the fingerprint in either the ZINC (xꢀaxis) or ZINClick diversity
subsets (yꢀaxis). To simplify the graphical representation, “null” is represented by “0.1”.
To investigate whether any ZINClick compounds can act against protein targets in vitro, a
systematic search within the 16 million compounds was carried out using the openꢀaccess
ChEMBL v.16 databases.47 This database contains published data regarding the biological
activity (without discrimination between active or inactive compounds) of approximately 1.6
ACS Paragon Plus Environment
15

Journal of Chemical Information and Modeling
Page 16 of 37
1
2
3
4
5
6
7
8
9
million small molecules (1.3 million compounds are reported as a chemical structure). In
total, 320 1,4ꢀdisubstituted triazoles are present, demonstrating the potential for finding novel
bioactive compounds using ZINClick; 320 triazoles represent only 0.002% of the ZINClick
database, confirming that ZINClick represents an underexplored region of chemical space.
Finally, the presence of nonspecific frequent hitters within screening libraries was
evaluated; this problem is commonly associated with false positives during screening
campaigns48. The compound collections provided for virtual screening often include
molecules that contain chemically reactive groups or other undesirable functionalities that
may interfere with the HTS detection techniques; these compounds are often referred to as
PAINS (panꢀassay interfering substances) or frequent hitters. We applied the PAINS
substructure filters published by Baell and Holloway41 to flag any compounds present in the
ZINClick database with PAINS structural motifs. ZINClick contains only eight different
PAINS scaffolds (Supporting Information Table S3 and Figure S5). In total, 1,281,422 (8%)
compounds have at least one PAINS substructure. In the current version of ZINClick, these
molecules were annotated but not removed, warning the operator while allowing him/her
complete control of the use of the ZINClick database.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
3.1. Availability and future directions. The ZINClick database is freely available online
via our website (http://www.symech.it/ZINClick). The current database is version 13; the
number refers to the year of release (January 2013). The database is updated with azides and
alkynes up the end of December 2012. A new version will be released every January.
The large size of the ZINClick database restricts users from downloading the entire
database in 3D multiꢀconformer format, but all 16 million structures and numerous subsets
(drugꢀlike, leadꢀlike, fragmentꢀlike, diversity and 10% subset) are freely available in the
canonical SMILES and/or sdf formats (minimized 3D structure). The files can be
ACS Paragon Plus Environment
16

Page 17 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
downloaded in a compressed format; they are also immediately available for docking or
virtual screening studies. Currently, a basic search functionality (by ZINClick id, physical
properties and chemical structure) is available for searching only the 10% subset.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 5. Detailed view of the ZINClick webpage showing the available subsets.
4. PRACTICAL USE – CASE STUDIES
Although, it is beyond of the scope of this work to test and compare the performances of
commonly used programs associated with virtual screening, we replicated some recent
studies to illustrate the potential utility of the ZINClick database.
4.1. Identification of NAADP analogues. NAADP (nicotinic acid adenine dinucleotide
phosphate) is a Ca2+ꢀreleasing second messenger that acts in various organisms ranging from
plants to mammals. Reportedly, this molecule interacts through an independent receptor that
differs from the others involved in Ca2+ release from storage. Furthermore, classical chemical
agents able to induce Ca2+ release do not affect NAADPꢀinduced Ca2+ release. Therefore, a
small organic molecule able to mimic NAADP is an interesting opportunity for medicinal
chemists. In 2009, Naylor et al. identified an NAADP antagonist through a threeꢀdimensional
shape and electrostatic mapꢀbased virtual screening using the ZINC database.49 After
following the reported protocol, we performed a threeꢀdimensional shape comparison
ACS Paragon Plus Environment
17

Journal of Chemical Information and Modeling
Page 18 of 37
1
2
3
4
5
6
7
8
9
between NAADP and the molecules in the ZINClick library using ROCS50 to calculate an
overlay of the two molecules. The compounds selected via ROCS were ranked using their
shape Tanimoto score. The top 500 ‘hits’ from ZINClick had a Tanimoto score between 0.69
and 0.77; in comparison, the values from ZINC were 0.66 and 0.73. We reꢀranked the top 500
compounds for similarity to NAADP based on the electrostatic Tanimoto score while using
the EON program.51 The top 500 ‘hits’ had an electrostatic Tanimoto score between 0.00 and
0.64, whereas the values of the original paper were ꢀ0.31 and 0.85. We selected the top 10
Nrdꢀclick hits (NAADP discovered by ROCS) and the top 15 Nedꢀclick hits (NAADP
discovered by EON). The details regarding the selected hits are reported in Supporting
Information Table S5. Remarkably, the best compounds selected by ROCS show a shape
Tanimoto score comparable to or even better than the best compounds reported by Naylor,
while the electrostatic Tanimoto scores are lower. The best compound was Nedꢀclickꢀ1, as
depicted in Figure 6, as it showed the best overlay with NAADP. This compound was never
reported and can be prepared in four synthetic steps (Scheme 1).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 6. The topꢀscored compounds in the case studies. (a) Twoꢀdimensional chemical
structures of NAADP and Nedꢀclickꢀ1. (b) Overlay of NAADP (green carbons) and Nedꢀ
clickꢀ1 (cyan carbons).
ACS Paragon Plus Environment
18

Page 19 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Scheme 1. Synthesis of compound Nedꢀclickꢀ1. Reagents and conditions: a) EDC, DMF, 3 h,
r.t.;52 b) NaH, DMF, 2 h at 80 °C than 4 h at 100 °C;52 c) tꢀBuONO, Me3SiN3, acetonitrile, 0
°C; 2 h, r.t.53
4.2. Identification of Glycogen Synthase Kinase 3β Inhibitors. The second case study
uses ZINClick to identify novel glycogen synthase kinase 3β (GSKꢀ3β) inhibitors. GSKꢀ3β is
a major protein kinase that is involved in the regulation of glucose metabolism, attracting
significant attention as a therapeutic target. In a recent study conducted by Osolodkin et al.,54
a crossꢀdocking study was used to identify the Xꢀray structure and combination of constraints
leading to the best reproduction of the binding positions for different inhibitors. While
following their inputs, the crystal structure of Glycogen Synthase Kinase 3β (PDB id: 1R0E)
was used, and two constraints were applied due the presence of hydrogen bonds between the
ligand and hinge residues (Asp133, backbone carbonyl oxygen atom, and Val135, backbone
ACS Paragon Plus Environment
19

Journal of Chemical Information and Modeling
Page 20 of 37
1
2
3
4
5
6
7
8
9
amide hydrogen atom). FRED55 software was used with the Chemgauss3 scoring function.
The details for the 10 best hits (Fꢀclick) are reported in Supporting Information Table S6. The
predicted binding pose of the bestꢀscoring ligand is plausible (Figure 7), and the binding
affinities show be ꢀ15.18 kcal/mol. This compound has never been reported and can be
prepared in two synthetic steps (Scheme 2). Studies to validate these two hypotheses are
ongoing in our laboratory, and the results will be reported in due course.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 7. The best predicted GSKꢀ3β inhibitor. (a) Twoꢀdimensional chemical structures of
DFN and Fꢀclickꢀ1. (b) Docked structure of Fꢀclickꢀ1 (pink sticks) in the GSKꢀ3β binding
site (green cartoon), with the crystal structure DFN superposed as cyan lines.
ACS Paragon Plus Environment
20

Page 21 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Scheme 2. Synthesis of compound Fꢀclickꢀ1. Reagents and conditions: a) triphosgene, TEA,
CH2Cl2, 20 min, 120 °C;56 b) DBU, CuI, (S)ꢀProline, DMSO, 20 min, 120 °C;56 c) NaOH,
NaN3, H2O, EtOH, r.t.; 2 h, 110 °C.57
5. CONCLUSIONS
As discussed above, one problem that cannot be underestimated with virtual screening, is
that most molecules present in several databases are not new, and therefore, many research
groups can discover the same molecules. Furthermore, the chemical inaccessibility of some
ACS Paragon Plus Environment
21

Journal of Chemical Information and Modeling
Page 22 of 37
1
2
3
4
5
6
active compounds is limited due to their long and/or difficult synthesis, frustrating the
operator.
7
8
9
In this manuscript, we have constructed a database containing 16 million 1,4ꢀdisubstituted
triazoles that are new, patentable and synthetically feasible. The chemistry used to create the
required azides and alkynes has already been published, and the final click chemistry reaction
is robust and tolerant of numerous functional groups, failing only in rare cases.16d Finally, we
used our ZINClick database with two previously published cases for comparison.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
This
database
is
freely
downloadable
online
via
our
website
(http://www.symech.it/ZINClick), and it will be enriched every year with novel azides and
alkynes reported in the literature. We encourage others to bring azides and alkynes (MW<
500 Da, synthesizable using no more than three synthetic steps) to our attention. We hope this
novel database can inspire drug hunters and find routine use by medicinal and computational
chemists searching for novel hits for their biological targets. Even if only one successful case
is reported, we could consider our work successful.
ASSOCIATED CONTENT
Supporting Information. The supporting information includes tables listing the molecular
descriptors considered in the present work, a table listing the PAINS structural motifs that
were present in ZINClick and figures showing the distribution of molecular descriptors for all
the reference databases. This material is available free of charge via the Internet at
http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
ACS Paragon Plus Environment
22

Page 23 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
*Phone: +39.0321.375.753, Fax: +39.0321.375.821, Eꢀmail: alberto.massarotti@unipmn.it.,
*Phone: +39.0321.375.857, Fax: +39.0321.375.821, Eꢀmail: giancesare.tron@unipmn.it.
Notes
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The authors declare no competing financial interests.
ACKNOWLEDGMENTS
A.M. and G.S. gratefully acknowledge OpenEye Scientific Software Inc. for providing an
academic license for their software packages. Financial support from the Università del
Piemonte Orientale (Italy) and FIRB 2012 “Infiammazione e cancro: approcci innovativi
basati su nanotecnologie” MIUR Italy is gratefully acknowledged. The authors thank NASA
and The Hubble Heritage Team (STScI) for the free use of the picture adapted for use in the
TCG (http://hubblesite.org/gallery/album/galaxy/pr1999041a/).
REFERENCES
(1) a) Walters, W. P.; Stahl, M. T.; Murcko, M. A. Virtual screening—an overview. Drug
Discov. Today 1998, 3, 160ꢀ178; b) Shoichet, B. K. Virtual screening of chemical
libraries. Nature 2004, 432, 862ꢀ865.
(2) Scior, T.; Bender, A.; Tresadern, G.; MedinaꢀFranco, J. L.; MartinezꢀMayorga, K.;
Langer, T.; CuanaloꢀContreras, K.; Agrafiotis, D. K. Recognizing pitfalls in virtual
screening: a critical review. J. Chem. Inf. Model 2012, 52, 867ꢀ881.
(3) a) Good, A. C.; Oprea, T. I. Optimization of CAMD techniques 3. Virtual screening
enrichment studies: a help or hindrance in tool selection? J. Comput. Aided Mol. Des.
2008, 22, 169ꢀ178; b) Schneider, G. Virtual screening: an endless staircase? Nat. Rev.
Drug Discov. 2010, 9, 273ꢀ276.
(4) a) Alvarez, J.; Shoichet, B. Virtual Screening in Drug Discovery. 1 ed.; CRC Press:
Boca Raton, FL, 2005; p 470; b) Sotriffer, C. Virtual Screening: Principles,
Challenges, and Practical Guidelines. 1 ed.; WileyꢀVCH: Weinheim, DE, 2011; Vol.
48, p 519.
(5) Chemistry
Databases
and
Search
Services
on
the
Web.
http://cactus.nci.nih.gov/links/chem_www.html (accessed 31ꢀ08ꢀ2013).
ACS Paragon Plus Environment
23

Journal of Chemical Information and Modeling
Page 24 of 37
1
2
3
4
5
6
7
8
9
(6) Irwin, J. J.; Shoichet, B. K. ZINCꢀꢀa free database of commercially available
compounds for virtual screening. J. Chem. Inf. Model 2005, 45, 177ꢀ182.
(7) Irwin, J. J.; Sterling, T.; Mysinger, M. M.; Bolstad, E. S.; Coleman, R. G. ZINC: A
Free Tool to Discover Chemistry for Biology. J. Chem. Inf. Model 2012, 52, 1757–
1768.
(8) Available
Chemicals
Directory
(ACD).
http://www.akosgmbh.de/Symyx/software/databases/acd.htm
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(9) Dorschner, K. V.; Toomey, D.; Brennan, M. P.; Heinemann, T.; Duffy, F. J.; Nolan, K.
B.; Cox, D.; Adamo, M. F.; Chubb, A. J. TIN ꢀ A Combinatorial Compound Collection
of Synthetically Feasible Multicomponent Synthesis Products. J. Chem. Inf. Model
2011, 51, 986ꢀ995.
(10) Tron, G. C. Off the Beaten Track: The Use of Secondary Amines in the Ugi Reaction.
Eu. J. Org. Chem. 2013, 2013, 1849ꢀ1859.
(11) a) Fink, T.; Reymond, J. L. Virtual exploration of the chemical universe up to 11 atoms
of C, N, O, F: assembly of 26.4 million structures (110.9 million stereoisomers) and
analysis for new ring systems, stereochemistry, physicochemical properties, compound
classes, and drug discovery. J. Chem. Inf. Model 2007, 47, 342ꢀ353; b) Blum, L. C.;
Reymond, J. L. 970 million druglike small molecules for virtual screening in the
chemical universe database GDBꢀ13. J. Am. Chem. Soc. 2009, 131, 8732ꢀ8733; c)
Ruddigkeit, L.; van Deursen, R.; Blum, L. C.; Reymond, J. L. Enumeration of 166
billion organic small molecules in the chemical universe database GDBꢀ17. J. Chem.
Inf. Model. 2012, 52, 2864ꢀ2875.
(12) a) Carey, J. S.; Laffan, D.; Thomson, C.; Williams, M. T. Analysis of the reactions used
for the preparation of drug candidate molecules. Org. Biomol. Chem. 2006, 4, 2337ꢀ
2347; b) Roughley, S. D.; Jordan, A. M. The medicinal chemist's toolbox: an analysis
of reactions used in the pursuit of drug candidates. J. Med. Chem. 2011, 54, 3451ꢀ3479.
(13) Fullbeck, M.; Michalsky, E.; Dunkel, M.; Preissner, R. Natural products: sources and
databases. Nat. Prod. Rep. 2006, 23, 347ꢀ356.
(14) Chuprina, A.; Lukin, O.; Demoiseaux, R.; Buzko, A.; Shivanyuk, A. Drugꢀ and leadꢀ
likeness, target class, and molecular diversity analysis of 7.9 million commercially
available organic compounds provided by 29 suppliers. J. Chem. Inf. Model 2010, 50,
470ꢀ479.
(15) Kolb, H. C.; Finn, M. G.; Sharpless, K. B. Click Chemistry: Diverse Chemical Function
from a Few Good Reactions. Angew. Chem. Int. Ed. Engl. 2001, 40, 2004ꢀ2021.
(16) a) Rostovtsev, V. V.; Green, L. G.; Fokin, V. V.; Sharpless, K. B. A stepwise huisgen
cycloaddition process: copper(I)ꢀcatalyzed regioselective "ligation" of azides and
terminal alkynes. Angew. Chem. Int. Ed. Engl. 2002, 41, 2596ꢀ2599; b) Tornoe, C. W.;
Christensen, C.; Meldal, M. Peptidotriazoles on solid phase: [1,2,3]ꢀtriazoles by
regiospecific copper(i)ꢀcatalyzed 1,3ꢀdipolar cycloadditions of terminal alkynes to
azides. J. Org. Chem. 2002, 67, 3057ꢀ3064; c) Wu, P.; Fokin, V. V. Catalytic Azide–
Alkyne Cycloaddition: Reactivity and Applications. Aldrichimica Acta 2007, 40, 7ꢀ17;
d) Bock, V. D.; Hiemstra, H.; van Maarseveen, J. H. CuIꢀCatalyzed Alkyne–Azide
“Click” Cycloadditions from a Mechanistic and Synthetic Perspective. Eu. J. Org.
Chem. 2006, 2006, 51ꢀ68; e) Meldal, M.; Tornoe, C. W. Cuꢀcatalyzed azideꢀalkyne
cycloaddition. Chem. Rev. 2008, 108, 2952ꢀ3015; f) Hein, J. E.; Fokin, V. V. Copperꢀ
catalyzed azideꢀalkyne cycloaddition (CuAAC) and beyond: new reactivity of copper(I)
acetylides. Chem. Soc. Rev. 2010, 39, 1302ꢀ1315.
(17) Sharpless, K. B. Secret life of enzymes: An aggressive strategy for drug discovery.
Abstract of Papers, 229th ACS NAtional Meeting, American Chemical Society: San
Diego, CA, United States, 2005; pp ORGNꢀ611.
ACS Paragon Plus Environment
24

Page 25 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
(18) Rydzewski, R. M. Real World Drug Discovery. 1 ed.; Elsevier Ltd.: 2008.
(19) Hein, C. D.; Liu, X. M.; Wang, D. Click chemistry, a powerful tool for pharmaceutical
sciences. Pharm Res 2008, 25, 2216ꢀ2230.
(20) a) Kolb, H. C.; Sharpless, K. B. The growing impact of click chemistry on drug
discovery. Drug Discov. Today 2003, 8, 1128ꢀ1137; b) Moses, J. E.; Moorhouse, A. D.
The growing applications of click chemistry. Chem. Soc. Rev. 2007, 36, 1249ꢀ1262; c)
Moorhouse, A. D.; Moses, J. E. Click chemistry and medicinal chemistry: a case of
"cycloꢀaddiction". ChemMedChem 2008, 3, 715ꢀ723; d) Tron, G. C.; Pirali, T.;
Billington, R. A.; Canonico, P. L.; Sorba, G.; Genazzani, A. A. Click chemistry
reactions in medicinal chemistry: applications of the 1,3ꢀdipolar cycloaddition between
azides and alkynes. Med. Res. Rev. 2008, 28, 278ꢀ308; e) Agalave, S. G.; Maujan, S.
R.; Pore, V. S. Click chemistry: 1,2,3ꢀtriazoles as pharmacophores. Chem. Asian J.
2011, 6, 2696ꢀ2718; f) Thirumurugan, P.; Matosiuk, D.; Jozwiak, K. Click chemistry
for drug development and diverse chemicalꢀbiology applications. Chem. Rev. 2013,
113, 4905ꢀ4979.
(21) Reactor, version 6.0.4; ChemAxon, 2013. http://www.chemaxon.com
(22) Suite 2012: Maestro, version 9.3; Schrödinger, LLC, New York, NY, 2012.
(23) Molecular Operating Environment (MOE), 2012.10; Chemical Computing Group Inc.,
1010 Sherbooke St. West, Suite #910, Montreal, QC, Canada, H3A 2R7, 2012.
(24) a) Leete, E. The Biosynthesis of the Alkaloids of Colchicum. III. The Incorporation of
Phenylalanineꢀ2ꢀC14 into Colchicine and Demecolcine. J. Am. Chem. Soc. 1963, 85,
3666ꢀ3669; b) Douguet, D. eꢀLEA3D: a computationalꢀaided drug design web server.
Nucleic Acids Res. 2010, 38, W615ꢀW621.
(25) Durrant, J. D.; McCammon, J. A. AutoClickChem: click chemistry in silico. PLoS
Comput. Biol. 2012, 8, e1002397.
(26) R Core Team. R: A Language and Environment for Statistical Computing, version
3.0.1; R Foundation for Statistical Computing, Vienna, Austria, 2013. http://www.Rꢀ
project.org
(27) The PyMOL Molecular Graphics System, version 1.3; Schrödinger LLC, 2010.
(28) SciFinder®. http://scifinder.cas.org
(29) a) TubertꢀBrohman, I. Perl and chemistry. The Perl Journal 2004, 8, 3ꢀ5; b) Tubertꢀ
Brohman, I. PerlMol—Perl modules for molecular chemistry, 2004.
http://www.perlmol.org
(30) a) Hummler, E.; Hansmann, I. Preferential nondisjunction of specific bivalents in
oocytes from Djungarian hamsters (Phodopus sungorus) following colchicine
treatment. Cytogenet. Cell Genet. 1985, 39, 161ꢀ167; b) Hawkins, P. C. D.; Skillman,
A. G.; Warren, G. L.; Ellingson, B. A.; Stahl, M. T. Conformer generation with
OMEGA: algorithm and validation using high quality structures from the Protein
Databank and Cambridge Structural Database. J. Chem. Inf. Model 2010, 50, 572ꢀ584;
c) Hawkins, P. C. D.; Nicholls, A. Conformer generation with OMEGA: learning from
the data set and the analysis of failures. J. Chem. Inf. Model 2012, 52, 2919ꢀ2936.
(31) Mailhes, J. B.; Marchetti, F. The influence of postovulatory ageing on the retardation of
mouse oocyte maturation and chromosome segregation induced by vinblastine.
Mutagenesis 1994, 9, 541ꢀ545.
(32) Knox, C.; Law, V.; Jewison, T.; Liu, P.; Ly, S.; Frolkis, A.; Pon, A.; Banco, K.; Mak,
C.; Neveu, V.; Djoumbou, Y.; Eisner, R.; Guo, A. C.; Wishart, D. S. DrugBank 3.0: a
comprehensive resource for 'omics' research on drugs. Nucleic Acids Res. 2011, 39,
D1035ꢀ1041.
(33) Lipinski, C. A. Drugꢀlike properties and the causes of poor solubility and poor
permeability. J. Pharmacol. Toxicol. Methods 2000, 44, 235ꢀ249.
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment
25

Journal of Chemical Information and Modeling
Page 26 of 37
1
2
3
4
5
6
7
8
9
(34) Teague, S. J.; Davis, A. M.; Leeson, P. D.; Oprea, T. The Design of Leadlike
Combinatorial Libraries. Angew. Chem. Int. Ed. Engl. 1999, 38, 3743ꢀ3748.
(35) Carr, R. A.; Congreve, M.; Murray, C. W.; Rees, D. C. Fragmentꢀbased lead discovery:
leads by design. Drug Discov. Today 2005, 10, 987ꢀ992.
(36) chemfp, version 1.1; Dalke Scientific, Sweden, 2013. http://chemfp.com
(37) OEChem TK, Python version 1.7.7; OpenEye Scientific Software, Santa Fe, NM.
http://www.eyesopen.com
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(38) GraphSim TK, Python version 2.0.1; OpenEye Scientific Software, Santa Fe, NM.
http://www.eyesopen.com
(39) Magli, M. C.; Gianaroli, L.; Ferraretti, A. P. Chromosomal abnormalities in embryos.
Mol. Cell Endocrinol. 2001, 183 Suppl 1, S29ꢀ34.
(40) Voigt, J. H.; Bienfait, B.; Wang, S.; Nicklaus, M. C. Comparison of the NCI open
database with seven large chemical structural databases. J. Chem. Inf. Comput. Sci.
2001, 41, 702ꢀ712.
(41) Baell, J. B.; Holloway, G. A. New substructure filters for removal of pan assay
interference compounds (PAINS) from screening libraries and for their exclusion in
bioassays. J. Med. Chem. 2010, 53, 2719ꢀ2740.
(42) Guha, R. PAINS Substructure Filters as SMARTS. http://blog.rguha.net/?p=850
(accessed 04ꢀ07ꢀ2013).
(43) Bendsen, E.; Byskov, A. G.; Andersen, C. Y.; Westergaard, L. G. Number of germ cells
and somatic cells in human fetal ovaries during the first weeks after sex differentiation.
Hum. Reprod. 2006, 21, 30ꢀ35.
(44) Two reasons drove us to use this cutꢀoff: a) the possibility to expand the chemical
complexity of the starting materials; b) to allow the operator to identify triazoles with
similar features to the reference molecule under study. See case study 1.
(45) Bender, A.; Glen, R. C. Molecular similarity: a key technique in molecular informatics.
Org. Biomol. Chem. 2004, 2, 3204ꢀ3218.
(46) a) Martin, Y. C.; Kofron, J. L.; Traphagen, L. M. Do structurally similar molecules
have similar biological activity? J. Med. Chem. 2002, 45, 4350ꢀ4358; b) Willett, P.;
Barnard, J. M.; Downs, G. M. Chemical Similarity Searching. J. Chem. Inf. Comp. Sci.
1998, 38, 983ꢀ996.
(47) Overington, J. ChEMBL. An interview with John Overington, team leader,
chemogenomics at the European Bioinformatics Institute Outstation of the European
Molecular Biology Laboratory (EMBLꢀEBI). Interview by Wendy A. Warr. J. Comput.
Aided. Mol. Des. 2009, 23, 195ꢀ198.
(48) a) Sink, R.; Gobec, S.; Pecar, S.; Zega, A. False positives in the early stages of drug
discovery. Curr. Med. Chem. 2010, 17, 4231ꢀ4255; b) Baell, J. B. Observations on
screeningꢀbased research and some concerning trends in the literature. Fut. Med. Chem.
2010, 2, 1529ꢀ1546.
(49) Naylor, E.; Arredouani, A.; Vasudevan, S. R.; Lewis, A. M.; Parkesh, R.; Mizote, A.;
Rosen, D.; Thomas, J. M.; Izumi, M.; Ganesan, A.; Galione, A.; Churchill, G. C.
Identification of a chemical probe for NAADP by virtual screening. Nat. Chem. Biol.
2009, 5, 220ꢀ226.
(50) a) ROCS, version 2.4.1; OpenEye Scientific Software, Santa Fe, NM.
http://www.eyesopen.com; b) Hawkins, P. C. D.; Skillman, A. G.; Nicholls, A.
Comparison of shapeꢀmatching and docking as virtual screening tools. J. Med. Chem.
2007, 50, 74ꢀ82.
(51) a) EON, version 2.0.1; OpenEye Scientific Software, Santa Fe, NM.
http://www.eyesopen.com; b) Muchmore, S. W.; Souers, A. J.; AkritopoulouꢀZanze, I.
The use of threeꢀdimensional shape and electrostatic similarity searching in the
ACS Paragon Plus Environment
26

Page 27 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
identification of a melaninꢀconcentrating hormone receptor 1 antagonist. Chem. Biol.
Drug. Des. 2006, 67, 174ꢀ176.
(52) Kubota, M.; Sakaguchi, H.; Kandoh, Y. Preparation of naphthyridinecarboxamides for
plant disease control. WO2009093640 (A1) 2009.
7
(53) Barral, K.; Moorhouse, A. D.; Moses, J. E. Efficient conversion of aromatic amines into
azides: a oneꢀpot synthesis of triazole linkages. Org. Lett. 2007, 9, 1809ꢀ1811.
(54) Osolodkin, D. I.; Palyulin, V. A.; Zefirov, N. S. Structureꢀbased virtual screening of
glycogen synthase kinase 3β inhibitors: analysis of scoring functions applied to large
true actives and decoy sets. Chem. Biol. Drug. Des. 2011, 78, 378ꢀ390.
(55) a) Magnuson, T.; Debrot, S.; Dimpfl, J.; Zweig, A.; Zamora, T.; Epstein, C. J. The early
lethality of autosomal monosomy in the mouse. J. Exp. Zool. 1985, 236, 353ꢀ360; b)
McGann, M. FRED pose prediction and virtual screening accuracy. J. Chem. Inf.
Model. 2011, 51, 578ꢀ596.
(56) Li, Z.; Sun, H.; Jiang, H.; Liu, H. Copperꢀcatalyzed intramolecular cyclization to Nꢀ
substituted 1,3ꢀdihydrobenzimidazolꢀ2ꢀones. Org. Lett. 2008, 10, 3263ꢀ3266.
(57) Jia, Z.; Zhu, Q. 'Click' assembly of selective inhibitors for MAOꢀA. Bioorg. Med.
Chem. Lett. 2010, 20, 6222ꢀ6225.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment
27

Journal of Chemical Information and Modeling
Page 28 of 37
1
2
3
4
Table of Contents Graphic
5
6
7
8
9
0000
ZINClick: a database of 16 million novel, patentable and readily
synthesizable 1,4ꢀdisubstituted triazoles
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
Alberto Massarotti,* Angelo Brunco, Giovanni Sorba, and Gian Cesare
Tron*
For Table of Contents Use Only
ACS Paragon Plus Environment
28

Page 29 of 37
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. Histogram of the Lipinski violations as percentages of the ZINClick (orange), ZINC (cyan), ZINC
“Natural product” (green) and DrugBank (yellow) databases. The details are reported in Supporting
Information Table S2.
115x80mm (300 x 300 DPI)
ACS Paragon Plus Environment