Optimizing TRPM4 inhibitors in the MHFP6 chemical space

Article abstract of DOI:10.1016/j.ejmech.2019.01.048

We recently reported 4-chloro-2-(2-chlorophenoxy)acetamido)benzoic acid (CBA) as the first potent inhibitor of TRPM4, a cation channel implicated in cardiac diseases and prostate cancer. Herein we report a structure-activity relationship (SAR) study of CBA resulting in two new potent analogs. To design and interpret our SAR we used interactive color-coded 3D-maps representing similarities between compounds calculated with MHFP6 (MinHash fingerprint up to six bonds), a new molecular fingerprint outperforming other fingerprints in benchmarking virtual screening studies. We further illustrate the general applicability of our method by visualizing the structural diversity of active compounds from benchmarking sets in relation to decoy molecules and to drugs. MHFP6 chemical space 3D-maps might be generally helpful in designing, interpreting and communicating the results of SAR studies. The modified WebMolCS is accessible at http://gdb.unibe.ch and the code is available at https://github.com/reymond-group/webMolCS for off-line use.

Full text of DOI:10.1016/j.ejmech.2019.01.048

Accepted Manuscript
Optimizing TRPM4 inhibitors in the MHFP6 chemical space
Clémence Delalande, Mahendra Awale, Matthias Rubin, Daniel Probst, Lijo C.
Ozhathil, Jürg Gertsch, Hugues Abriel, Jean-Louis Reymond
PII:
S0223-5234(19)30060-1
DOI:
https://doi.org/10.1016/j.ejmech.2019.01.048
EJMECH 11055
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 24 October 2018
Revised Date: 18 December 2018
Accepted Date: 19 January 2019
Please cite this article as: Clé. Delalande, M. Awale, M. Rubin, D. Probst, L.C. Ozhathil, Jü. Gertsch, H.
Abriel, J.-L. Reymond, Optimizing TRPM4 inhibitors in the MHFP6 chemical space, European Journal of
Medicinal Chemistry (2019), doi: https://doi.org/10.1016/j.ejmech.2019.01.048.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

1
ACCEPTED MANUSCRIPT
Optimizing TRPM4 Inhibitors in the MHFP6 Chemical Space
Clémence Delalande,a) Mahendra Awale,a) Matthias Rubin,b) Daniel Probst,a) Lijo C. Ozhathil,b)
Jürg Gertsch,b) Hugues Abrielb) and Jean-Louis Reymonda)*
a) Department of Chemistry and Biochemistry, National Center for Competence in Research NCCR
TransCure, University of Berne, Freiestrasse 3, 3012 Bern, Switzerland
b) Institute of Biochemistry and Molecular Medicine, National Center of Competence in Research
NCCR TransCure, University of Bern, Bühlstrasse 28, 3012 Bern, Switzerland
e-mail: jean-louis.reymond@dcb.unibe.ch
Abstract
We recently reported 4-chloro-2-(2-chlorophenoxy)acetamido)benzoic acid (CBA) as the first
potent inhibitor of TRPM4, a cation channel implicated in cardiac diseases and prostate cancer.
Herein we report a structure-activity relationship (SAR) study of CBA resulting in two new potent
analogs. To design and interpret our SAR we used interactive color-coded 3D-maps representing
similarities between compounds calculated with MHFP6 (MinHash fingerprint up to six bonds), a
new molecular fingerprint outperforming other fingerprints in benchmarking virtual screening
studies. We further illustrate the general applicability of our method by visualizing the structural
diversity of active compounds from benchmarking sets in relation to decoy molecules and to drugs.
MHFP6 chemical space 3D-maps might be generally helpful in designing, interpreting and
communicating the results of SAR studies. The modified WebMolCS is accessible at
http://gdb.unibe.ch and the code is available at https://github.com/reymond-group/webMolCS for
off-line use.
Keywords: Ion channels, chemical space, structure-activity relationships, cheminformatics, virtual
screening

2
ACCEPTED MANUSCRIPT
Introduction
We recently reported the halogenated anthranilic acid derivative 4-chloro-2-(2-
chlorophenoxy)acetamido)benzoic acid (CBA) as the first potent and selective inhibitor of
TRPM4,1 a calcium-activated nonselective cation channel expressed in many tissues and implicated
in cardiac diseases2 and prostate cancer.3 CBA shows reversible inhibition of TRPM4
overexpressed in HEK293 cells, and blocks TRPM4-mediated currents in human prostate cancer
cells. CBA also acts as chemical chaperone to rescue the expression and activity of a loss-of-
function TRPM4 genetic variant. We discovered CBA by selecting analogs of the known but weak
and unselective inhibitors 9-phenanthrol,4 glibenclamide and flufenamic acid from commercial
catalogs using a previously validated 3D-shape and pharmacophore similarity method,5, 6 and
screening them in a cell-based sodium uptake assay. The initial screening campaign also identified
4-chloro-2-(2-(4-chloro-2-methylphenoxy)propanamido) benzoic acid (LBA) with a lactic acid
linker as potent TRPM4 inhibitor (Figure 1).
Analysis of the structural diversity of compounds used for the discovery of CBA using
WebMolCS,7 a web-based application for creating interactive 3D-representations of chemical space
(3D-maps) available at http://gdb.unibe.ch, had shown that CBA resided in a region of chemical
space which only contained few commercial compounds and could only be explored by additional
chemical synthesis. Initial structure-activity relationship (SAR) studies focused on variations of the
2-chlorophenoxy group, which uncovered 4-chloro-2-(1-naphthyloxyacetamido)benzoic acid
(NBA) with an even stronger activity than CBA.1 We also prepared several isosteres of the carboxyl
group, however these efforts did not yield any active analogs.
Herein we report an additional SAR study leading to the discovery of two new TRPM4
inhibitors (8 and 9) with potency comparable to CBA. To design and interpret this new SAR study
we used a modified version of WebMolCS displaying chemical space 3D-maps calculated using
MHFP6 (MinHash fingerprint up to six bonds), a new molecular fingerprint using the principle of

3
ACCEPTED MANUSCRIPT
circular substructures,8, 9 and recently reported by our group.10 Molecular fingerprints are
multidimensional vector representations of molecular structures used to quantify similarities
between molecules.11-14 Such comparisons can be calculated extremely rapidly, and can uncover
similarities that are not easily perceived by visual inspection of molecules, although they are mostly
valid in the perspective of drug design, as assessed by benchmarking studies.15 In these
benchmarking studies MHFP6 outperforms ECFP4 (extended connectivity fingerprint up to four
bonds),16 which was up to now the best fingerprint across a large number of different molecular
fingerprints. We were therefore interested to test the suitability of MHFP6 for designing SAR
studies.
Figure 1. Structure and activity of TRPM4 inhibitors.
In the present report, we first modified our web-based application WebMolCS by enabling 3D-
representations of the MHFP6 chemical space. We then computed a 3D-map of the MHFP6

4
ACCEPTED MANUSCRIPT
chemical space representing the initial set of compounds used to discover our TRPM4 inhibitor
CBA and used the map to select new analogs. We discuss the synthesis and testing of these analogs
leading to the discovery of the new inhibitors 8 and 9, and how MHFP6 chemical space 3D-maps
provide an overview of the SAR study. Finally, we show that MHFP6 chemical space 3D-maps can
be useful in a general context by illustrating structural differences among target specific compound
series from the directory of useful decoys.17 MHFP6 chemical space 3D-maps might be generally
helpful in designing, interpreting and communicating the results of SAR studies by providing
overviews that are difficult to gain from R-group variation tables. The modified WebMolCS is
accessible at http://gdb.unibe.ch and the code is available at https://github.com/reymond-
group/webMolCS for off-line use.
Results and Discussion
Extending WebMolCS with MHFP6 chemicals space 3D-maps
In WebMolCS the user can load a series of up to 5,000 molecules written as a list of SMILES
annotated with a name and optionally a property value such as an activity, and then obtain an
interactive 3D-chemical space representation of the dataset (3D-map) within a few minutes. In these
3D-maps, molecules are represented as spheres colored-coded from red to blue either by similarity
rank to the first molecule in the list (FP-Rank), the rank in the list of molecules (List-Rank), or the
user-defined property value (ActualVals). The spheres, whose corresponding molecular structure is
shown on mouse-over, are positioned at coordinates obtained by projecting a high-dimensional
chemical space defined by a molecular fingerprint into 3D by dimensionality reduction. The spheres
can be optionally connected with each other by a minimum spanning tree to facilitate the perception
of nearest neighbor relationships, and the PC axes can be optionally displayed.
WebMolCS uses principal component analysis (PCA) as the dimensionality reduction
method to project simple fingerprint spaces such as MQN (Molecular Quantum Numbers, 42D),18, 19
SMIfp (SMILES fingerprint, 34D)20 or APfp (Atom-Pair fingerprint, 20D),21 because PCA can be

5
ACCEPTED MANUSCRIPT
parallelized and is fast even for millions of molecules, as recently illustrated with our web-based
visualization application Fearun.22 However, PCA is not possible with MHFP6 because this
fingerprint is a set rather than a vector of bit indices, and can only serve to compute similarities
between two molecules.10 We therefore implemented similarity mapping to project the MHFP6
chemical space into 3D. In similarity mapping, one first computes an N-dimensional similarity
space whose N dimensions are the similarity values to N randomly selected reference molecules,
and then performs a PCA of this similarity fingerprint to generate a 2D- or 3D-representation (see
methods for details).23-25 The approach is interesting because the calculation of similarity maps is
much faster than other dimensionality reduction methods for visualizing chemical space,26-28 and is
therefore applicable to very large datasets.
Remarkably, similarity mapping allows to map chemical spaces for high-dimensional
fingerprints that do not project well by PCA, including ECFP4 and MHFP6. In such cases, the
choice of reference molecules for the similarity fingerprint is critical. In our preliminary studies
with MHFP6, we found that similarity maps calculated using only actives as references, as
implemented previously in 2D-maps of various sets of actives in relation to ChEMBL,25 are
straightforward to interpret because the first three principal components, which define the three
dimensions of the map, usually cover over 80 % of data variance. In many cases PC1 itself covers
50-70% of data variance, reflecting the fact that most series of active compounds feature small and
incremental structural variations. We therefore modified WebMolCS to read an optional fourth data
column in the SMILES list of molecules called “satellite option”, which allows one to mark if a
molecule should be used as reference (value = 1) or not (value = 0) for constructing a similarity
map.
The creation of MHFP6 chemical space 3D-maps, as implemented in the modified version
of WebMolCS including the satellite option to mark actives as references, is presented in detail in
the methods section. Its use is illustrated in the SAR study and analysis of benchmarking sets
presented below.

6
ACCEPTED MANUSCRIPT
MHFP6 chemicals space maps of the initial TRPM4 screening set and SAR design
Using the modified WebMolCS described above, we computed a MHFP6 chemical space 3D-map
featuring all compounds investigated during the discovery of our TRPM4 inhibitor CBA using only
active compounds as references. The map color-coded by compound activity values and oriented
with the main PC1 axis pointing to the right is shown in Figure 2. In this image, CBA appears at
front right. The insert on top left shows the same map color coded by MHFP6 similarity rank to
CBA, which illustrates that the PC1 axis reflects structural differences separating 9-phenanthrol and
glibenclamide (far left) from CBA and analogs (far right). CBA and analogs are anthranilic acid
amides related to flufenamic acid, which appears in the center of the map.
Figure 2. MHFP6 chemical space map of initial screening set leading to CBA. Main map: map color-coded by activity
(blue = inactive, red = most active). Variance covered: red axis PC1 = 79 %, blue axis PC2 = 8 %, green axis PC3 = 3
%. Insert upper left: the same view color-coded by MHFP6 similarity rank (rank 1 = red, last rank = blue). The
interactive map is available at
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1544785130174&fp=MHFP
The 3D-map shows that our initial SAR covered a diversity of analogs of the 2-chlorophenoxy
group that resulted in the identification of NBA (2-(1-naphthyloxyacetamido)-4-chloro-benzoic

7
ACCEPTED MANUSCRIPT
acid) as a potent analog of CBA (Figure 2, upper right). The map also shows that only few active
compounds are found among analogs featuring variations of the anthranilic acid moiety (Figure 2,
lower center). Among variations of the 4-chloro substituent, only the 4-bromo analog of CBA
showed good activity, while other substituents (F, Me, CF3) at that position led to inactive analogs.
Furthermore, any replacement of the carboxyl group abolished the activity including isosteres such
as tetrazole (Figure 2, lower left).
Despite of this apparent lack of activity, the total number of anthranilic acid variations
surveyed in our initial study was still low, suggesting to consider additional cases accessible either
from commercial building blocks or by synthetic modifications. We set out to explore the chemical
space around the bromo analog of CBA by substituting the bromo substituent with larger groups.
Furthermore, despite of the inactivity of carboxylic isosteres, we aimed for an additional attempt
with an N-sulfonyl carboxamide analog, a type of carboxyl group replacement that can be used to
improve pharmacokinetic properties.29, 30 Note that the map also points to a relative lack of analogs
around the lactic acid linker analogs LBA (top left of the image in Figure 4), however we decided
not to pursue LBA further because both enantiomers had similar activities and the compound was
prone to racemization.1
Synthesis
Alkylation of 2-chlorophenol with tert-butyl chloroacetate yielded ester 1, which was deprotected
by acidic treatment to 2, and coupled with commercially available anthranilic acid esters to yield
intermediates 3, 5 and 7. Basic hydrolysis of the methyl esters using lithium hydroxide yielded the
4-cyano (4), 4,5-dichloro (6) and 6-methyl (8) analogs of CBA as final products in moderate yields
due to the competing hydrolysis of the anthranilic amide bond to both acidic and basic ester
deprotection conditions (Scheme 1).

8
ACCEPTED MANUSCRIPT
Scheme 1. Synthesis of CBA analogs from commercially available anthranilic esters.
Further derivatives were prepared by modification of CBA and close analogs. First, CBA itself was
converted by direct coupling to the corresponding N-cyclopropylsulfonyl carboxamide 9 (Scheme
2). Furthermore, palladium catalyzed (Suzuki or Stille) coupling of the aromatic bromide in methyl
ester 10 with boronic acid or tributyl tin derivates of five-membered aromatic heterocycles gave
ester intermediates 11, 13, 16 and 18, which were subsequently hydrolyzed as above to yield
carboxylic acids 12, 14, 17 and 19. Product 15 was obtained directly as the free carboxylic acid
during the palladium-coupling step. All final products showed >95% purity (analytical HPLC).

9
ACCEPTED MANUSCRIPT
a) cyclopropanesulfonamide,
Cl
Cl
Oxyma, DIC, collidine,
DCM, RT, 2h30, 17%
Cl
O
Cl
O
O
O
N
N
H
O
H
CO2H
S
O
N
O
H
CBA
9
N
O
Br
a) R-B(OH)2, Pd(PPh3)4, K2CO3,
dioxane/H2O/MeOH,
Cl
O
90°C, 2h-20h, 10-72%
O
Cl
O
N
H
O
CO2Me
N
H
10
CO2R
e) LiOH,
dioxane,
RT, 1h, 33%
11 (R = Me & Et)
12 (R = H)
S
Cl
O
N
N
N
N
O
N
H
CO2R
Cl
O
Cl
O
13 (R = Me)
14 (R = H)
e) LiOH,
dioxane,
RT, 1h, 23%
O
O
N
H
N
H
CO2H
CO2H
15 (87:13)
N
O
N
S
a) R-Sn(Bu)3,
Pd(PPh3)4,
THF, 65°C,
15-18h, 34-86%
Cl
O
Cl
O
O
O
N
N
H
H
10
CO2R
CO2R
e) LiOH,
dioxane,
RT, 1h, 24%
e) LiOH,
dioxane,
RT, 1h, 14%
16 (R = Me)
17 (R = H)
18 (R = Me)
19 (R = H)
Scheme 2. Synthesis of CBA analogs by derivatization.
TRPM4 inhibition
We measured the activity of the new analogs using a previously developed cell-based assay.1 In this
assay, HEK293 cells overexpressing TRPM4 are first depleted of Na+, and then loaded with the Na+
sensitive dye ANG-II using the acetoxylmethyl (ACM) protected cell permeable precursor.
Intracellular esterases cleave the ACM groups to release the active sensor, which remains trapped
inside cells. Exposure to 140 mM Na+ and 10 µM ionomycin to trigger Ca2+ influx and activation of
TRPM4, results in Na+ influx into the cells that can be followed in real time by the increase in

10
ACCEPTED MANUSCRIPT
ANG-II fluorescence (Figure 3). Screening of the nine newly synthesized analogs at 10 µM showed
significant inhibition with all compounds except 7 and 14. We then determined IC50 values for the
seven active inhibitors and performed duplicate determinations for the two most active inhibitors 8
and 9 (Table 1).
Figure 3. Fluorescent cell-based assay to determine TRPM4 modulation, measured in TRPM4 overexpressing cells.
Red line: negative control (140 mM Na+, no ionomycin), black line: positive control (140 mM Na+, 10 µM ionomycin),
green line: inhibitor (140 mM Na+, 10 µM ionomycin, 25 µM 9-phenanthrol).
Table 1. Inhibition of TRPM4 by CBA analogs.a)
Cl
O
R1
O
N
H
R2
Cpd.
6
7
R1
R2
IC50
3.6 µM
30% inhibition at 10 µM
1.3 ± 0.1 µM
4-cyano
4,5-dichloro
6-methyl
COOH
COOH
COOH
8
4-chloro
CONHSO2cProp 2.8 ± 0.4 µM
COOH
9
3,5-dimethylisoxazol-4-yl
thiophen-3-yl
1-methyl-1H-pyrazol-4-yl &
1-methyl-1H-pyrazol-5-yl
oxazol-2-yl
14
15
16
no inhibition up to 100 µM
COOH
COOH
5.9 µM
9.1 µM
COOH
COOH
4.9 µM
4.7 µM
19
20
thiazol-2-yl
a) Compound inhibition was measured in a cell-based assay. Concentration-response curve was constructed for all active
inhibitors except 7 and the values were fitted with Hill equation to extrapolate IC50 values. Duplicates were performed
for 8, 9 and data are given as mean ± sem.

11
ACCEPTED MANUSCRIPT
Interpreting SAR results using a MHFP6 chemical space 3D-map
To gain an overview of the overall SAR study, we computed a new MHFP6 chemical space 3D-
map using WebMolCS including all molecules used in the previous map combined with the newly
designed analogs, again using only active compounds for the similarity map calculation. This new
map is similar to the map in Figure 2 and is shown in a similar layout color-coded by compound
activity (Figure 4). In this view, CBA analogs with variations of the 2-chlororphenoxy group appear
at front right and CBA analogs with variation of the anthranilic acid appear in the lower central part
of the image. The insert at upper right shows the same map color-coded by MHFP6 similarity rank
to CBA, which illustrates the gradient of similarity along the PC1 axis from CBA itself at front
right to 9-phenanthrol at far left.
Figure 4. MHFP6 chemical space map of the overall SAR of TRPM4 analogs. Main map: map color-coded by activity
(blue = inactive, red = most active). Variance covered: red axis PC1 = 78 %, blue axis PC2 = 8 %, green axis PC3 = 4
%. Insert upper right: the same view color-coded by MHFP6 similarity rank (rank 1 = red, last rank = blue). The
interactive map is available at
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1544605933001&fp=MHFP

12
ACCEPTED MANUSCRIPT
In the map in Figure 4, all newly synthesized analogs are connected by the spanning tree between
the cyano analog 6, the 6-methyl analog 8 and the N-sulfonylcarboxamide 9. The map thus
illustrates that the newly synthesized analogs improved coverage of chemical space in that region of
the map, as suggested by their design. The map furthermore shows that exploring chemical space
around the bromo analog (red dot connected to 6) only uncovered less potent analogs such as the
cyano derivative 6 and the heteroaromatic analogs 15, 16, 19 and 20 (shown as green dots around
compound 8). The map also points to the fact that TRPM4 inhibition is very sensitive to the precise
substitution pattern. For example, the activity of CBA is preserved in its 2,5-dichlorophenoxy
analog but disappears in the 2,4-dichlorphenoxy analog despite of their high similarity indicated by
their nearest neighbor relationship in the 3D-map.
Visualizing actives, decoys and drugs in the MHFP6 chemical space
MHFP6 chemical space 3D-maps might be generally useful to inspect series of target specific
actives. This is illustrated here with datasets from the directory of useful decoys, DUD-E,17 which
groups series of molecules active on the same target mixed with decoy molecule with similar
overall physico-chemical properties. MHFP6 chemical space 3D-maps show that series of actives
from DUD-E typically consist in three to four different series of analogs, each featuring molecules
with a specific substructure pattern and a range of different substituents, as illustrated here for
inhibitors of thymidine kinase (kith), catechol-O-methyl transferase (comt), focal adhesion kinase
(fak1), and farnesyl diphosphate synthase (fpps) (Figure 5).
In all cases, the decoys are well separated from the actives, reflecting the performance of
MHFP6 in virtual screening benchmarks. The strong separation between actives and decoys in
Figure 5 contrasts to poor separation between actives and inactives in our TRPM4 inhibition study
(Figures 2 and 4). This difference reflects the fact that decoy molecules in benchmarking sets are
molecules taken at random in a diverse pool of compounds only preserving the overall physico-

13
ACCEPTED MANUSCRIPT
chemical properties of the actives but not their detailed substructures, while our TRPM4 series
features many compounds sharing common scaffolds and small changes in substituents.
Figure 5. Visualization of benchmarking sets from DUD-E using MHFP6 similarity maps in WebMolCS, color coded
as actives (red) or decoys (blue). Interactive 3D-maps can be accessed at the indicated weblinks. (a) Thymidine kinase
(kith), PC1 = 70 % , PC2 = 9 %, PC3 = 5 %,
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1544775928296&fp=MHFP. (b) catechol O-methyl
transferase (comt), PC1 = 72 % , PC2 = 11 %, PC3 = 3 %,
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1544776368253&fp=MHFP. (c) focal adhesion kinase
1 (fak1), PC1 = 70 % , PC2 = 8 %, PC3 = 5 %,
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1544776863167&fp=MHFP. (d) farnesyl diphosphate
synthase (fpps), PC1 = 43 % , PC2 = 12 %, PC3 = 11 %,
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1544777667633&fp=MHFP.
MHFP6 chemical space 3D-maps can also be constructed using actives from one, two, three of four
different targets as references. In this case we added approximately the 2,200 approved drugs listed
in DrugBank for comparison, as illustrated here for the four target specific series discussed above
(Figure 6).31 Each series of actives appears as an elongated group separated from the approved

14
ACCEPTED MANUSCRIPT
drugs, which cluster in the center of the map because none of them is considered as reference in this
display. The 3D-maps highlight the main structural patterns in each target specific series. For focal
adhesion kinase (fak1), this motif is a pyrrolopyridine, for thymidine kinase (kith), it is a thymidine-
like substructure, for farnesyl diphosphate synthase (fpps) a diphosphonate group, and for catechol-
O-methyl transferase (comt) a catechol. These maps show that our MHFP6 chemical space 3D-
maps also provide useful differentiation between compound series with different bioactivities.
Figure 6. Visualization of active sets from DUD-E in combination with approved drugs from DrugBank using MHFP6
similarity maps in WebMolCS, color coded by set (actual value option in controls). (a) green: Fak1 (focal adhesion
kinase 1), red: kith (thymidine kinase), blue: DrugBank,
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1539954046166&fp=MHFP. (b) green: kith
(thymidine kinase), red: fpps (farnesyl diphosphate synthase), cyan: Fak1 (focal adhesion kinase 1), blue: DrugBank,
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1539951847766&fp=MHFP. (c) green: kith (thymidine
kinase), red: comt (catechol O-methyl transferase), cyan: Fak1 (focal adhesion kinase 1), blue: DrugBank,
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1539951512218&fp=MHFP. (d) green: kith
(thymidine kinase), red: fpps (farnesyl diphosphate synthase), cyan: Fak1 (focal adhesion kinase 1), yellow: comt
(catechol O-methyl transferase), blue: DrugBank,
http://gdbtools.unibe.ch:8080/webMolCS/yourSIM.html?jobID=1539952304206&fp=MHFP.

15
ACCEPTED MANUSCRIPT
Conclusion
In summary, we modified our web-based application WebMolCS to enable 3D-maps displaying
similarity relationships between compounds calculated with MHFP6, a recently reported molecular
fingerprint, which outperforms ECFP4 in virtual screening benchmarking studies. We then used our
modified WebMolCS to compute a MHFP6 chemical space 3D-map displaying a set of compounds
used to discover CBA, a potent inhibitor of the cation channel TRPM4. The map helped us to
design a new series of analogs, which we synthesized and tested for TRPM4 inhibition, leading to
the discovery of compounds 8 and 9 as two new CBA analogs with comparable potencies. We also
found that MHFP6 chemical space 3D-maps provide insightful representations of various series of
target specific actives and their relationship to decoy molecules and to drugs. We propose MHFP6
chemical space 3D-maps as an alternative to R-group variation tables for designing, interpreting and
communicating the results of SAR studies.
Methods
Cheminformatics
Loading datasets. The WebMolCS webpage at http://gdb.unibe.ch takes as input dataset an
annotated list of molecules in which each line of the file represents a different molecule and
comprises the following entries: 1) the SMILES of the molecule; 2) the name or code for the
molecule; 3) a value for color coding, which may be a number representing activity as pIC50 or a
number corresponding to a selective subset; 4) a satellite selection value, set to 1 if the compound
should be considered as reference for the similarity map calculation, and 0 if it should not be
considered as reference.
MHFP6 fingerprint calculation. Besides the series of fingerprints already present, the modified
WebMolCS web application now includes an option for computing similarity maps in the MHFP6

16
ACCEPTED MANUSCRIPT
chemical space, which uses the 128D version of MHFP6. Calculating the MHFP6 fingerprint of a
molecule comprises the following operations: 1) write the SMILES of all circular substructures up
to a diameter of six bonds around each atom; 2) convert each SMILES to a value using a Hash
function; 3) Apply the MinHash procedure to this set of values. For the 128D MHFP6 considered
here, MinHash applies 128 different Hash functions to each value in the primary set, yielding 128
new values for each substructure. For each of the 128 different hash functions, only the smallest of
the hashed values is kept. These minimum hashed values are collected to form a secondary set of
values, which is the MHFP6 fingerprint. The method is probabilistic in that each substructure only
has a certain probability of being considered in the final set of secondary values generated by the
MinHash procedure. The resulting 128D MHFP6 is a set of 128 values. The similarity between two
molecules is calculated from their MHFP6 fingerprints as the ratio of the number of common values
to the total number of different values in the two sets (Jaccard index).
Similarity map calculation. To produce the MHFP6 similarity map, WebMolCS selects as
references either 50 molecules taken at regular interval across the entire dataset, or the subset of
molecules with the satellite option value set to 1. The server then performs the following operations:
1) compute the MHFP6 fingerprint similarity value of each compound in the list to each of the
reference compounds; 2) calculate the PCA on the resulting similarity fingerprint; 3) construct a
3D-representation of the dataset using the first three principal components (PCs) as coordinates, in
which each molecule appears as a sphere positioned at its (PC1, PC2, PC3) coordinates, and the
molecular structure is shown on mouse-over. The color-code for the spheres can be selected in the
menu, and the spheres can be connected with each other by lines representing the minimum
spanning tree, which connects nearest neighbors in the 3D-space. The entire computation from the
input dataset to the MHFP6 similarity map requires approximately 2.2 minutes of server time for a
set of 1,000 molecules.

17
ACCEPTED MANUSCRIPT
Availability. The WebMolCS webpage at http://gdb.unibe.ch is freely accessible for use. The
source code for the application and instructions to setup WebMolCS in a dockerized container are
available on Github at https://github.com/reymond-group/webMolCS.
Synthesis
Reagents and solvents were purchased from commercial sources and were used without further
purification. MilliQ deionised water and HPLC-grade acetonitrile were employed for RP-HPLC
solvents. Synthesis were followed both by Alugram Xtra Sil G/UV254 with detection of the
compounds by UV at 214 nm or by UPLC-MS. RP-UPLC spectra were obtained using a Dionex
Ultimate 3000 RSLC System and a Dionex Acclaim RSLC 120 column. Three methods were
performed as follows: A gradient from 0.05 % TFA to 9:1 acetonitrile/water 0.05 % TFA during 2.2
min then 2.8 min of 9:1 acetonitrile/water 0.05 % TFA isocratic (method A). A gradient from 0.05
% TFA to 9:1 acetonitrile/water 0.05 % TFA during 4.5 min then 2.5 min of 9:1 acetonitrile/water
0.05 % TFA isocratic (method B). A gradient from 0.05 % TFA to 9:1 acetonitrile/water 0.05 %
TFA during 7.5 min then 2.5 min of 9:1 acetonitrile/water 0.05 % TFA isocratic (method C). Low-
resolution mass spectra were conducted on a Thermo Scientific LCQ Fleet by electron spray
ionization (ESI) in the positive mode. Automated column chromatography purification was
executed on an Interchim Puriflash 430 system on RediSepRf prepacked columns. Preparative RP-
HPLC was performed on a Waters Prep LC4000 Chromatography system using a Reprospher 100
column from Dr. Maisch GmbH with a flow of 40 mL/min and two lines of solvents: a 0.1 % TFA
and a 9:1 acetonitrile/water 0.1 % TFA. Compounds were analyzed by UPLC-MS and their purity
assigned by analytical RP-UPLC. NMR spectra were recorded on: BRUKER AC 300 for 1H-300
MHz; 13C-75 MHz; 19F-282 MHz measurements or BRUKER AC 400 for 1H-400 MHz; 13C-101
MHz measurements. Chemical shifts are given in parts per million (δ) referenced to TMS (δ = 0.00
ppm 1H-, 13C-NMR). Coupling constants are given in Hertz. High resolution mass spectra
measurements were performed by the “Service of Mass Spectrometry” of the Department of

18
ACCEPTED MANUSCRIPT
Chemistry and Biochemistry at the University of Bern on a Thermo Scientific LTQ OritrapXL
(group of PD Dr. Stefan Schürch).
tert-Butyl 2-(2-chlorophenoxy)acetate (1). Potassium carbonate (46.7 mmol; 6.45 g) was added to
a solution of 2-chlorophenol (23.3 mmol; 2.38 mL) in DMF (10 mL) and stirred for 10 min at RT.
The tert-butyl chloroacetate (25.7 mmol; 3.67 mL) was added then the mixture was heated to 80°C
for 2 h until the reaction was complete. After full conversion, the hot mixture was poured into a
mixture of crushed ice and water (~50 mL). The mixture was stirred for 30 min then the precipitate
was filtered off and dried to afford a white solid (5.50 g; 97%), which was used without further
treatment. 1H NMR (300 MHz, CDCl3) δ 7.38 (dd, J = 7.9, 1.6 Hz, 1H), 7.19 (ddd, J = 8.2, 7.6, 1.6
Hz, 1H), 6.93 (td, J = 7.7, 1.4 Hz, 1H), 6.82 (dd, J = 8.2, 1.3 Hz, 1H), 4.60 (s, 2H), 1.48 (s, 9H).
2-(2-Chlorophenoxy)acetic acid (2). The ester 1 (22.7 mmol; 5.50 g) was dissolved in
trifluoroacetic acid (22.7 mL) and stirred at RT for 1 hour and a half until the reaction was
complete. The crude was concentrated under reduced pressure. The obtained solid was taken back
with DCM several times then evaporated to dryness to afford a white solid (4.20 g; 99%), which
was used without further treatment. 1H NMR (300 MHz, DMSO) δ 13.10 (s, 1H), 7.43 (dd, J = 7.9,
1.6 Hz, 1H), 7.27 (ddd, J = 8.5, 7.5, 1.6 Hz, 1H), 7.06 - 6.92 (m, 2H), 4.80 (s, 2H).
Methyl 2-(2-(2-chlorophenoxy)acetamido)-4-cyanobenzoate (3). Acid 2 (2.20 mmol; 0.42 g) was
dissolved in 12 mL toluene and SOCl2 (82.3 mmol; 6.00 mL) was added. The mixture was refluxed
overnight, and the solvent was removed on high vacuum to give the crude acyl chloride. This
product was dissolved in a small amount of THF and added to a solution of methyl 2-amino-4-
cyanobenzoate (2.00 mmol; 0.35 g) and K2CO3 (4.00 mmol; 0.55 g) in THF (29 ml) stirred for 15
min at RT. The reaction mixture was stirred at RT for 4 h until the reaction was complete. The
reaction was diluted with water (30 mL) and extracted with EtOAc (3 x 30 mL). The combined
organics were washed with brine, dried with Na2SO4 and evaporated to dryness. The crude product
was recrystallized in acetonitrile to give white crystals (0.55 g; 80%). 1H NMR (400 MHz, CDCl3)

19
ACCEPTED MANUSCRIPT
δ 11.90 (s, 1H), 9.16 (d, J = 1.4 Hz, 1H), 8.13 (d, J = 8.2 Hz, 1H), 7.42 (ddd, J = 12.3, 8.0, 1.5 Hz,
2H), 7.25 (td, J = 8.0, 1.6 Hz, 1H), 7.05 – 6.96 (m, 2H), 4.72 (s, 2H), 3.95 (s, 3H). 13C NMR (101
MHz, CDCl3) δ 167.45, 166.71, 153.10, 140.55, 131.72, 130.96, 128.10, 126.38, 124.51, 123.86,
123.53, 119.61, 117.77 (d, J = 5.3 Hz), 114.82, 69.11, 53.13. HRMS m/z [M + H]+ calcd for
C17H14ClN2O4 345.0637, found: 345.0627. mp: 159.6-160.1°C.
2-(2-(2-chlorophenoxy)acetamido)-4-cyanobenzoic acid (4). A 1.4 M aqueous solution of lithium
hydroxide (9.24 mmol; 6.60 mL) was added to a solution of 3 (1.54 mmol; 0.53 g) in 13 mL
dioxane and stirred at RT for 1 h until the reaction was complete. The reaction mixture was
quenched with 3.1 mL of a 1M HCl solution and concentrated under reduced pressure. The crude
was first filtrated on a silica pad with ethyl acetate acidified with acetic acid then purified by
automated flash chromatography on 25 g of silica gel (gradient from 98 % 2 % DCM/MeOH to 95
% 10% DCM/MeOH) two times. The impure fractions obtained from the column were taken back
in water and the formed precipitate was filtered and then triturated in ethanol three times to give a
pure white solid (0.004 g; 1 %).1H NMR (400 MHz, DMSO) δ 14.32 (s, 1H), 11.88 (s, 1H), 8.98 (d,
J = 1.2 Hz, 1H), 8.14 (d, J = 8.2 Hz, 1H), 7.66 (dd, J = 8.2, 1.4 Hz, 1H), 7.48 (dd, J = 7.9, 1.4 Hz,
1H), 7.35 – 7.28 (m, 1H), 7.23 – 7.17 (m, 1H), 7.07 – 7.00 (m, 1H), 4.91 (s, 2H). 13C NMR (101
MHz, DMSO) δ 167.94, 167.49, 152.65, 139.96, 132.25, 130.19, 128.35, 126.52, 123.20, 122.76,
121.74, 120.78, 117.89, 115.77, 114.47, 68.12. HRMS m/z [M + H]+ calcd for C16H12ClN2O4
331.0480, found: 331.0473. mp = 30°C. UPLC (method A): tR = 2.18 min., purity: 99 %.
Methyl 4,5-dichloro-2-(2-(2-chlorophenoxy)acetamido)benzoate (5). Acid 2 (1.10 mmol; 0.20 g)
was dissolved in 12 mL toluene and SOCl2 (82.3 mmol; 6.00 mL) was added. The mixture was
refluxed overnight, and the solvent was removed on high vacuum to give the crude acyl chloride.
This product was dissolved in a small amount of THF and added to a solution of methyl 2-amino-
4,5-dichloro-benzoate (1.00 mmol; 0.22 g) and K2CO3 (2.01 mmol; 0.28 g) in THF (13 ml) stirred
for 15 min at RT. The reaction mixture was stirred at RT for 4 h until the reaction was complete.
The reaction was diluted with water (30 mL) and extracted with EtOAc (3 x 30 mL). The combined

20
ACCEPTED MANUSCRIPT
organics were washed with brine, dried with Na2SO4 and evaporated to dryness to give a yellow
solid which was used without further purification (0.38 g; 97%). 1H NMR (300 MHz, CDCl3) δ
11.83 (s, 1H), 9.01 (s, 1H), 8.11 (s, 1H), 7.44 (dd, J = 7.8, 1.5 Hz, 1H), 7.23 (dd, J = 7.9, 1.5 Hz,
1H), 7.05 – 6.95 (m, 2H), 4.70 (s, 2H), 3.92 (s, 3H). 13C NMR (75 MHz, CDCl3) δ 167.06, 153.04,
139.04, 132.08, 130.83, 127.95, 126.92, 123.35, 122.57, 117.54, 115.84, 114.70, 69.04, 52.80. One
quaternary carbon was not observed and is expected to be close to the baseline. HRMS m/z [M +
Na]+ calcd for C16H12Cl3NO4Na 409.9724, found: 409.9716. mp: 158.2-160.4°C.
4,5-Dichloro-2-(2-(2-chlorophenoxy)acetamido)benzoic acid (6). A 1.4 M aqueous solution of
lithium hydroxide (5.82 mmol; 4.16 mL) was added to a solution of 5 (0.97 mmol; 0.38 g) in 12 mL
dioxane and stirred at RT for 1 h until the reaction was complete. The reaction mixture was
quenched with 1.9 mL of a 1M HCl solution and concentrated under reduced pressure. The crude
was taken back in 0.1 % TFA and the formed precipitate was filtered and washed with water twice.
The obtained precipitate was first taken back in HCl 1M two times and filtrated and then triturated
by MeOH to give a white solid (0,15 mg; 41%). 1H NMR (400 MHz, DMSO) δ 14.25 (s, 1H), 11.85
(s, 1H), 8.91 (s, 1H), 8.07 (s, 1H), 7.55 – 7.40 (m, 1H), 7.30 (s, 1H), 7.23 – 7.12 (m, 1H), 7.02 (s,
1H), 4.87 (s, 2H). 13C NMR (101 MHz, DMSO) δ 167.32, 152.62, 139.26, 136.30, 132.22, 130.17,
128.29, 124.95, 122.74, 121.79, 121.31, 117.12, 114.46, 68.14. HRMS m/z [M + Na]+ calcd for
C15H10Cl3NO4Na 395.9568, found: 395.9568. HRMS m/z [M - H]+ calcd for C15H9Cl3NO4
371.9603, found: 371.9613. mp: 267.6-268.6°C. UPLC (method A): tR = 2.75 min., purity: 98 %.
Methyl 2-(2-(2-chlorophenoxy)acetamido)-6-methylbenzoate (7). Acid 2 (1.10 mmol; 0.20 g)
was dissolved in 12 mL toluene and SOCl2 (82.3 mmol; 6.00 mL) was added. The mixture was
refluxed overnight, and the solvent was removed on high vacuum to give the crude acyl chloride.
This product was dissolved in a small amount of THF and added to a solution of methyl 2-amino-6-
methyl-benzoate (1.00 mmol; 0.17 g) and K2CO3 (2.00 mmol; 0.28 g) in THF (17 ml) stirred for 15
min at RT. The reaction mixture was stirred at RT for 3 h until the reaction was complete. The
reaction was diluted with water (30 mL) and extracted with EtOAc (3 x 30 mL). The combined

21
ACCEPTED MANUSCRIPT
organics were washed with brine, dried with Na2SO4 and evaporated to dryness to give a brown oil
which was used without further purification (0.39 g; 99%). 1H NMR (300 MHz, CDCl3) δ 10.07 (s,
1H), 8.17 (d, J = 8.3 Hz, 1H), 7.44 (dd, J = 7.9, 1.6 Hz, 1H), 7.37 (t, J = 8.0 Hz, 1H), 7.24 (dd, J =
8.1, 1.6 Hz, 1H), 7.06 – 7.00 (m, 2H), 7.00 – 6.92 (m, 1H), 4.68 (s, 2H), 3.85 (s, 3H), 2.43 (s, 3H).
13C NMR (75 MHz, CDCl3) δ 168.47, 166.39, 153.12, 138.68, 136.37, 131.50, 130.82, 128.15,
127.49, 123.59, 123.27, 122.25, 120.25, 114.45, 68.93, 52.32, 21.88. HRMS m/z [M + H]+ calcd for
C17H17ClNO4 334.0841, found: 334.0845.
2-(2-(2-chlorophenoxy)acetamido)-6-methylbenzoic acid (8). A 1.4 M aqueous solution of
lithium hydroxide (6.93 mmol; 4.95 mL) was added to a solution of 7 (1.16 mmol; 0.39 g) in 14 mL
dioxane and stirred at RT for 1 h until the reaction was complete. The reaction mixture was
quenched with 2.3 mL of a 1M HCl solution and concentrated under reduced pressure. The crude
was taken back in 0.1 % TFA and the formed precipitate was filtered and washed with MeOH
twice. The obtained precipitate was first triturated by tert-butylmethylether several times then DCM
several times to give a white solid (0,06 mg; 17%). 1H NMR (400 MHz, DMSO) δ 12.77 (s, 1H),
8.07 (d, J = 8.1 Hz, 1H), 7.45 (dd, J = 7.9, 1.6 Hz, 1H), 7.28 (ddd, J = 8.5, 7.5, 1.6 Hz, 1H), 7.13
(dd, J = 8.3, 1.3 Hz, 1H), 7.05 – 6.96 (m, 2H), 6.79 (d, J = 7.5 Hz, 1H), 4.70 (s, 2H), 2.39 (s, 3H).
13C NMR (101 MHz, DMSO) δ 169.21, 165.27, 153.48, 136.61, 136.38, 130.87, 130.04, 128.14,
126.29, 125.36, 122.35, 122.01, 116.93, 114.85, 68.77, 21.90. HRMS m/z [M + H]+ calcd for
C16H15ClNO4 320.0684, found: 320.0684. mp: 254.0-258.8°C. UPLC (method A): tR = 2.28 min.,
purity: 96 %.
4-Chloro-2-(2-(2-chlorophenoxy)acetamido)-N-(cyclopropylsulfonyl)benzamide (9). A solution
of CBA (0.29 mmol; 0.10 g), Oxyma (0.44 mmol; 0.06 g), DIC (0.44 mmol; 0.07 mL) and collidine
(0.44 mmol; 0.06 mL) was added to a solution of the cyclopropanesulfonamide (0.88 mmol; 0.11
mg) in DCM (4 mL). The reaction mixture was stirred at RT for 2h30 until the reaction was
complete. The reaction mixture was concentrated under reduced pressure and the obtained crude
was further purified by RP-HPLC (gradient from 0% to 80 % 9:1 acetonitrile/water 0.1 % TFA in

22
ACCEPTED MANUSCRIPT
40 minutes) to give a pure batch of the compound and two other impure fractions that were
recrystallized in ethyl acetate to give a white lyophilisate (0.02 g; 17%). 1H NMR (400 MHz,
DMSO) δ 11.94 (s, 1H), 8.76 (d, J = 2.1 Hz, 1H), 8.01 (d, J = 8.6 Hz, 1H), 7.48 (dd, J = 7.9, 1.6 Hz,
1H), 7.31 – 7.26 (m, 3H), 7.17 (ddd, J = 14.0, 8.3, 1.2 Hz, 2H), 7.03 (td, J = 7.7, 1.4 Hz, 2H), 4.89
(s, 2H), 2.91 (tt, J = 8.1, 4.8 Hz, 1H), 1.11 – 0.88 (m, 3H). 13C NMR (101 MHz, DMSO) δ 168.42,
167.36, 152.69, 140.99, 138.50, 132.93, 130.20, 128.35, 123.21, 122.74, 119.39, 115.27, 114.43,
114.28, 68.19, 30.39, 5.27. HRMS m/z [M + Na]+ calcd for C18H16Cl2N2O5SNa 465.0049, found:
465.0040. UPLC (method A): tR = 2.34 min., purity: 97 %.
Methyl 4-bromo-2-(2-(2-chlorophenoxy)acetamido)benzoate (10). Acid 2 (21.4 mmol; 4.00 g)
was dissolved in 16 mL toluene and SOCl2 (107 mmol; 7.82 mL) was added. The mixture was
refluxed overnight, and the solvent was removed on high vacuum to give the crude acyl chloride.
This product was dissolved in a small amount of THF and added to a solution of methyl 2-amino-4-
bromobenzoate (19.4 mmol; 4.48 g) and K2CO3 (38.9 mmol; 5.38 g) in THF (115 ml) stirred for 15
min at RT. The reaction mixture was stirred at RT overnight until the reaction was complete. The
reaction was diluted with water (30 mL) and extracted with EtOAc (3 x 30 mL). The combined
organics were washed with brine, dried with Na2SO4 and evaporated to dryness to give the pure
product as a brown solid (6.40 g; 83 %). 1H NMR (400 MHz, CDCl3) δ 11.88 (s, 1H), 9.02 (d, J =
1.9 Hz, 1H), 7.89 (d, J = 8.5 Hz, 1H), 7.44 (dd, J = 7.7, 1.5 Hz, 1H), 7.29 (dd, J = 8.5, 2.0 Hz, 1H),
7.23 (dd, J = 7.8, 1.5 Hz, 1H), 7.06 – 6.96 (m, 2H), 4.71 (s, 2H), 3.90 (s, 3H). 13C NMR (101 MHz,
CDCl3) δ 167.51, 167.21, 153.28, 141.04, 132.13, 130.94, 129.40, 128.06, 126.76, 123.96, 123.91,
123.41, 115.12, 114.84, 69.26, 52.67. HRMS m/z [M + Na]+ calcd for C16H13BrClNO4Na 419.9609,
found: 419.9601. mp: 127.1-130.2°C.
Methyl 2-(2-(2-chlorophenoxy)acetamido)-4-(3,5-dimethylisoxazol-4-yl)benzoate & Ethyl 2-(2-
(2-chlorophenoxy)acetamido)-4-(3,5-dimethylisoxazol-4-yl)benzoate (11). 10 (0.38 mmol; 0.15
g), (3,5-dimethylisoxazol-4-yl)boronic acid (0.56 mmol; 0.08 g), K2CO3 (1.13 mmol; 0.16 g) and
palladium tetrakis (0.02 mmol; 0.02 g) were dissolved in a mixture of dioxane/water/ethanol (2 mL;

23
ACCEPTED MANUSCRIPT
4:1:10) degased under argon. The reaction mixture was stirred at 80°C for 2h until the reaction was
complete. The reaction mixture was concentrated under reduced pressure and taken back in water
and diethyl ether. The organic phase was washed with brine, dried with Na2SO4 and concentrated
under reduced pressure. The obtained crude was further purified by automated flash
chromatography on 12 g of silica gel (gradient from 80 % 20 % Hexane/EtOAc to 100 % EtOAc) to
give a yellow oil corresponding to a 50/50 mixture of both esters (0.10 g; 63 %). 1H NMR (400
MHz, CDCl3) δ 11.97 (s, 1H), 11.89 (s, 1H), 8.76 (t, J = 1.4 Hz, 2H), 8.11 (dd, J = 8.2, 5.3 Hz, 2H),
7.44 (ddd, J = 8.0, 2.3, 1.7 Hz, 2H), 7.23 (dd, J = 5.6, 3.9 Hz, 1H), 7.06 (dt, J = 8.2, 1.8 Hz, 2H),
7.01 (tt, J = 8.0, 2.2 Hz, 4H), 4.72 (s, 4H), 4.39 (q, J = 7.1 Hz, 2H), 3.93 (s, 3H), 2.50 (d, J = 0.9
Hz, 6H), 2.36 (d, J = 0.7 Hz, 6H), 1.39 (t, J = 7.1 Hz, 3H). 13C NMR (101 MHz, CDCl3) δ 167.31,
167.24, 166.39, 153.35, 140.56 (d, J = 8.7 Hz), 136.84 (d, J = 14.7 Hz), 131.46 (d, J = 6.6 Hz),
130.94 (d, J = 1.7 Hz), 128.06 (d, J = 2.4 Hz), 123.94, 123.46 (dd, J = 17.7, 5.3 Hz), 121.37 (d, J =
7.3 Hz), 115.59, 115.35, 114.85 (d, J = 2.3 Hz), 69.29, 61.71, 52.62, 14.37, 12.05, 11.19.
2-(2-(2-chlorophenoxy)acetamido)-4-(3,5-dimethylisoxazol-4-yl)benzoic acid (12). A 1.4 M
aqueous solution of lithium hydroxide (1.40 mmol; 1.00 mL) was added to a solution of 11 (0.23
mmol; 0.10 g) in 2 mL dioxane and stirred at RT for 1 h until the reaction was complete. The
reaction mixture was quenched with 0.5 mL of a 1M HCl solution and concentrated under reduced
pressure. The crude was first filtrated on a silica pad with ethyl acetate acidified with acetic acid
then triturated in dry methanol and tert-butylmethylether to give a yellow solid (0.03 g; 33 %). 1H
NMR (400 MHz, DMSO) δ 8.61 (s, 1H), 8.06 (d, J = 8.1 Hz, 1H), 7.47 (dd, J = 7.9, 1.4 Hz, 1H),
7.34 – 7.24 (m, 1H), 7.15 (dd, J = 12.2, 4.5 Hz, 2H), 7.05 – 6.97 (m, 1H), 4.83 (s, 2H), 2.45 (s, 3H),
2.27 (s, 3H). 13C NMR (101 MHz, DMSO) δ 168.63, 166.69, 165.61, 157.94, 153.15, 140.22,
131.61, 130.14, 128.26, 122.67, 122.47, 121.84, 119.43, 115.49, 114.49, 68.48, 11.48, 10.62.
HRMS m/z [M + H]+ calcd for C20H18ClN2O5 401.0899, found: 401.0899. mp: 196.6-204.2°C.
UPLC (method A): tR = 2.48 min., purity: > 99 %.

24
ACCEPTED MANUSCRIPT
Methyl 2-(2-(2-chlorophenoxy)acetamido)-4-(thiophen-3-yl)benzoate (13). 10 (0.44 mmol; 0.18
g), 3-thienylboronic acid (0.66 mmol; 0.08 g), K2CO3 (1.32 mmol; 0.18 g) and palladium tetrakis
(0.02 mmol; 0.02 g) were dissolved in a mixture of dioxane/water/methanol (2 mL; 4:1:10) degased
under argon. The reaction mixture was stirred at 80°C for 2h until the reaction was complete. The
reaction mixture was concentrated under reduced pressure and taken back in water and diethyl
ether. The formed precipitate was filtered to give the pure product as a brown solid (0.13 g; 72 %).
1H NMR (400 MHz, CDCl3) δ 11.90 (s, 1H), 9.08 (d, J = 1.7 Hz, 1H), 8.06 (d, J = 8.3 Hz, 1H), 7.66
– 7.63 (m, 1H), 7.48 (dd, J = 5.0, 1.4 Hz, 1H), 7.44 (dd, J = 7.8, 1.5 Hz, 1H), 7.41 (dd, J = 5.1, 3.0
Hz, 1H), 7.38 (dd, J = 8.3, 1.8 Hz, 1H), 7.23 (dd, J = 7.9, 2.1 Hz, 1H), 7.06 – 6.95 (m, 2H), 4.73 (s,
2H), 3.91 (s, 3H). 13C NMR (101 MHz, CDCl3) δ 167.85, 167.21, 153.46, 141.56, 141.10, 140.81,
131.67, 130.92, 128.05, 126.74, 126.51, 123.98, 123.32, 122.83, 121.27, 118.68, 114.92, 114.87,
69.43, 52.45. HRMS m/z [M + Na]+ calcd for C20H16ClNO4SNa 424.0381, found: 424.0384. mp:
137.1-138.8°C.
2-(2-(2-chlorophenoxy)acetamido)-4-(thiophen-3-yl)benzoic acid (14). A 1.4 M aqueous solution
of lithium hydroxide (1.87 mmol; 1.33 mL) was added to a solution of 13 (0.31 mmol; 0.13 g) in
2.5 mL dioxane and stirred at RT for 2 h until the reaction was complete. The reaction mixture was
quenched with 0.6 mL of a 1M HCl solution and concentrated under reduced pressure. The crude
was first filtrated on a silica pad with ethyl acetate acidified with acetic acid then triturated in
ethanol to give a yellow solid (0.03 g; 23 %). 1H NMR (400 MHz, DMSO) δ 13.63 (s, 1H), 11.91
(s, 1H), 9.01 (d, J = 1.7 Hz, 1H), 8.02 (d, J = 8.3 Hz, 1H), 7.98 (dd, J = 2.9, 1.3 Hz, 1H), 7.72 (dd, J
= 5.0, 2.9 Hz, 1H), 7.58 – 7.51 (m, 2H), 7.48 (dd, J = 7.9, 1.6 Hz, 1H), 7.35 – 7.28 (m, 1H), 7.20
(dd, J = 8.3, 1.2 Hz, 1H), 7.03 (td, J = 7.7, 1.3 Hz, 1H), 4.89 (s, 2H). 13C NMR (101 MHz, DMSO)
δ 168.92, 167.04, 152.82, 140.61, 140.29, 140.24, 131.89, 130.20, 128.34, 127.86, 125.99, 123.39,
122.67, 121.78, 120.92, 117.23, 114.99, 114.41, 68.28. HRMS m/z [M + H]+ calcd for
C19H15ClNO4S 388.0405, found: 388.0404. mp: 240.5-268.1°C. UPLC (method A): tR = 2.64 min.,
purity: 99 %.

25
ACCEPTED MANUSCRIPT
2-(2-(2-chlorophenoxy)acetamido)-4-(1-methyl-1H-pyrazol-4-yl)benzoic acid & 2-(2-(2-
chlorophenoxy)acetamido)-4-(1-methyl-1H-pyrazol-5-yl)benzoic acid (15). 10 (0.63 mmol; 0.25
g), 1-methyl-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)pyrazole (0.94 mmol; 0.20 g), K2CO3
(1.88 mmol; 0.26 g) and palladium tetrakis (0.03 mmol; 0.04 g) were dissolved in a mixture of
dioxane/water/methanol (4 mL; 4:1:10) degased under argon. The reaction mixture was stirred at
80°C overnight until the reaction was complete. The reaction mixture was concentrated under
reduced pressure and taken back in water and diethyl ether. The formed precipitate corresponding to
the ester was filtered and the aqueous phase is freeze-dried. The obtained lyophilisate was filtrated
on a silica pad with ethyl acetate acidified with acetic acid then triturated in ethyl acetate and tert-
butylmethylether to give a yellow solid (0.02 g; 10 %; products ratio 87:13). 1H NMR (400 MHz,
DMSO) δ 15.33 (s, 1H), 8.66 (d, J = 1.7 Hz, 1H), 8.04 (s, 1H), 7.92 (d, J = 8.0 Hz, 1H), 7.73 (d, J =
0.5 Hz, 1H), 7.44 (dd, J = 7.9, 1.6 Hz, 1H), 7.31 – 7.24 (m, 1H), 7.13 (ddd, J = 9.5, 8.2, 1.5 Hz,
2H), 6.98 (td, J = 7.7, 1.4 Hz, 1H), 4.77 (s, 2H), 3.85 (s, 3H). 13C NMR (101 MHz, DMSO) δ
168.78, 165.75, 153.66, 140.65, 135.85, 133.56, 131.57, 130.01, 128.10, 127.82, 123.79, 122.08,
122.04, 121.84, 118.34, 114.72, 114.51, 68.70, 38.55. HRMS m/z [M + H]+ calcd for C19H17ClN3O4
386.0902, found: 386.0900. HRMS m/z [M + Na]+ calcd for C19H16ClN3O4Na 408.0722, found:
408.0717. UPLC (method A): tR = 2.10 & 2.48 min., purity: 99 %.
Methyl 2-(2-(2-chlorophenoxy)acetamido)-4-(oxazol-2-yl)benzoate (16). 10 (0.50 mmol; 0.20 g),
tributyl(oxazol-2-yl)stannane (1.05 mmol; 0.22 mL) and palladium tetrakis (0.40 mmol; 0.46 g)
were dissolved in THF (50 mL) degased under argon. The reaction mixture was stirred at 65°C
overnight until the reaction was complete. The reaction mixture was concentrated under reduced
pressure and taken back in EtOAc/DCM/Hexane. The formed precipitate was filtered, and the
filtrate was concentrated under reduced pressure. The obtained crude was further purified by
automated flash chromatography on 15 g of silica gel (isocratic eluent of 2:1 Hexane/EtOAc) to
give a brown solid (0.07 g; 34 %).1H NMR (400 MHz, CDCl3) δ 11.88 (s, 1H), 9.43 (d, J = 1.5 Hz,
1H), 8.14 (d, J = 8.3 Hz, 1H), 7.86 (dd, J = 8.3, 1.6 Hz, 1H), 7.78 (s, 1H), 7.44 (dd, J = 8.3, 1.5 Hz,

26
ACCEPTED MANUSCRIPT
1H), 7.30 (s, 1H), 7.24 (dd, J = 8.1, 1.3 Hz, 1H), 7.02 (d, J = 7.8 Hz, 2H), 4.75 (s, 2H), 3.93 (s, 3H).
13C NMR (101 MHz, CDCl3) δ 167.50, 167.18, 160.89, 153.35, 140.64, 139.68, 132.55, 131.68,
130.93, 129.18, 128.07, 123.90, 123.34, 121.19, 118.70, 117.55, 114.80, 69.30, 52.71. HRMS m/z
[M + H]+ calcd for C19H16ClN2O5 387.0742, found: 387.0745. mp: 145-146°C.
2-(2-(2-chlorophenoxy)acetamido)-4(oxazol-2-yl)benzoic acid (17). A 1.4 M aqueous solution of
lithium hydroxide (1.00 mmol; 0.72 mL) was added to a solution of 16 (0.17 mmol; 0.06 g) in 2 mL
dioxane and stirred at RT for 1 h until the reaction was complete. The reaction mixture was
quenched with 0.3 mL of a 1M HCl solution and concentrated under reduced pressure. The crude
was first filtrated on a silica pad with ethyl acetate acidified with acetic acid then triturated in
ethanol to give a first pure batch of the compound. The trituration filtrate was concentrated under
reduced pressure and recrystallized in H2O/ACN to give grey crystals that are gathered with the first
batch (0.02 g; 24 %). 1H NMR (400 MHz, DMSO) δ 12.05 (s, 1H), 9.35 (d, J = 1.5 Hz, 1H), 8.32
(s, 1H), 8.14 (d, J = 8.3 Hz, 1H), 7.78 (dd, J = 8.3, 1.6 Hz, 1H), 7.49 (d, J = 1.5 Hz, 1H), 7.47 (s,
1H), 7.35 – 7.28 (m, 1H), 7.20 (dd, J = 8.3, 1.1 Hz, 1H), 7.03 (td, J = 7.8, 1.3 Hz, 1H), 4.90 (s, 2H).
13C NMR (101 MHz, DMSO) δ 168.63, 167.14, 159.78, 152.80, 141.03, 140.41, 132.17, 131.20,
130.18, 129.04, 128.32, 122.69, 121.80, 120.30, 118.26, 117.04, 114.46, 68.31. HRMS m/z [M +
H]+ calcd for C18H14ClN2O5 373.0586, found: 373.0581. mp: 253.3-253.4°C. UPLC (method A): tR
= 2.17 min., purity: 96 %.
Methyl 2-(2-(2-chlorophenoxy)acetamido)-4-( thiazol-2-yl)benzoate (18). 10 (0.50 mmol; 0.20
g), tributyl(thiazol-2-yl)stannane (1.05 mmol; 0.33 mL) and palladium tetrakis (0.40 mmol; 0.46 g)
were dissolved in THF (50 mL) degased under argon. The reaction mixture was stirred at 65°C
overnight until the reaction was complete. The reaction mixture was concentrated under reduced
pressure and the obtained crude was further purified by automated flash chromatography on 50 g of
silica gel (isocratic eluent of 2:1 Hexane/EtOAc) to give a yellow solid (0.17 g; 86 %). 1H NMR
(400 MHz, CDCl3) δ 11.91 (s, 1H), 9.39 (d, J = 1.7 Hz, 1H), 8.13 (d, J = 8.3 Hz, 1H), 7.94 (d, J =
3.2 Hz, 1H), 7.84 (dd, J = 8.4, 1.8 Hz, 1H), 7.45 (dd, J = 6.3, 2.1 Hz, 2H), 7.25 – 7.22 (m, 1H), 7.01

27
ACCEPTED MANUSCRIPT
(dd, J = 12.0, 4.5 Hz, 2H), 4.75 (s, 2H), 3.93 (s, 3H). 13C NMR (101 MHz, CDCl3) δ 167.56,
167.28, 166.97, 153.39, 144.31, 140.81, 138.67, 131.85, 130.94, 128.07, 123.97, 123.37, 121.09,
120.54, 119.16, 117.21, 114.90, 69.36, 52.66. HRMS m/z [M + H]+ calcd for C19H16ClN2O4S
403.0514, found: 403.0506. HRMS m/z [M + Na]+ calcd for C19H15ClN2O4SNa 425.0333, found:
425.0324. mp: 71.4-71.5°C.
2-(2-(2-chlorophenoxy)acetamido)-4-(thiazol-2-yl)benzoic acid (19). A 1.4 M aqueous solution
of lithium hydroxide (2.51 mmol; 1.79 mL) was added to a solution of 18 (0.42 mmol; 0.17 g) in 4
mL dioxane and stirred at RT for 1 h until the reaction was complete. The reaction mixture was
quenched with 0.8 mL of a 1M HCl solution and concentrated under reduced pressure. The crude
was first filtrated on a silica pad with ethyl acetate acidified with acetic acid then recrystallized in
acetonitrile to give impure crystals that were triturated in tert-butylmethylether to give a yellow
solid (0.02 g; 14 %). 1H NMR (400 MHz, DMSO) δ 13.17 (s, 1H), 9.29 (d, J = 1.4 Hz, 1H), 8.09 (d,
J = 8.2 Hz, 1H), 7.99 (d, J = 3.2 Hz, 1H), 7.87 (d, J = 3.2 Hz, 1H), 7.69 (dd, J = 8.2, 1.6 Hz, 1H),
7.47 (dd, J = 7.9, 1.4 Hz, 1H), 7.33 – 7.26 (m, 1H), 7.18 (dd, J = 8.2, 0.9 Hz, 1H), 7.02 (td, J = 7.8,
1.2 Hz, 1H), 4.87 (s, 2H). 13C NMR (101 MHz, DMSO) δ 168.48, 166.86, 166.26, 153.09, 144.21,
140.58, 140.58, 136.16, 136.16, 132.14, 130.15, 128.28, 122.51, 121.83, 121.46, 120.18, 116.86,
114.49, 68.45. HRMS m/z [M + H]+ calcd for C18H14ClN2O4S 389.0357, found: 389.0364. mp:
228.1-228.3°C. UPLC (method A): tR = 2.30 min., purity: 95 %.
Acknowledgment. This work was supported financially by the Swiss National Science Foundation,
NCCR TransCure.
Competing interest statement. The authors declare not competing interests.

28
ACCEPTED MANUSCRIPT
References
1.
Ozhathil, L. C.; Delalande, C.; Bianchi, B.; Nemeth, G.; Kappel, S.; Thomet, U.; Ross-
Kaschitza, D.; Simonin, C.; Rubin, M.; Gertsch, J.; Lochner, M.; Peinelt, C.; Reymond, J. L.;
Abriel, H. Identification of Potent and Selective Small Molecule Inhibitors of the Cation Channel
Trpm4. Br. J. Pharmacol. 2018, 175, 2504-2519.
2.
Wang, C.; Naruse, K.; Takahashi, K. Role of the Trpm4 Channel in Cardiovascular
Physiology and Pathophysiology. Cells 2018, 7.
3.
Hantute-Ghesquier, A.; Haustrate, A.; Prevarskaya, N.; Lehen'kyi, V. Trpm Family
Channels in Cancer. Pharmaceuticals (Basel) 2018, 11, 58.
4.
Guinamard, R.; Hof, T.; Del Negro, C. A. The Trpm4 Channel Inhibitor 9-Phenanthrol. Br.
J. Pharmacol. 2014, 171, 1600-1613.
5.
Simonin, C.; Awale, M.; Brand, M.; van Deursen, R.; Schwartz, J.; Fine, M.; Kovacs, G.;
Häfliger, P.; Gyimesi, G.; Sithampari, A.; Charles, R.-P.; Hediger, M. A.; Reymond, J.-L.
Optimization of Trpv6 Calcium Channel Inhibitors Using a 3d Ligand-Based Virtual Screening
Method. Angew. Chem. Int. Ed. 2015, 54, 14748-14752.
6.
Kilchmann, F.; Marcaida, M. J.; Kotak, S.; Schick, T.; Boss, S. D.; Awale, M.; Gönczy, P.;
Reymond, J.-L. Discovery of a Selective Aurora a Kinase Inhibitor by Virtual Screening. J. Med.
Chem. 2016, 59, 7188-7211.
7.
Awale, M.; Probst, D.; Reymond, J. L. Webmolcs: A Web-Based Interface for Visualizing
Molecules in Three-Dimensional Chemical Spaces. J Chem Inf Model 2017, 57, 643-649.
8.
Morgan, H. L. The Generation of a Unique Machine Description for Chemical Structures-a
Technique Developed at Chemical Abstracts Service. J. Chem. Doc. 1965, 5, 107-113.
9.
Penny, R. H. A Connectivity Code for Use in Describing Chemical Structures. J. Chem.
Doc. 1965, 5, 113-117.
10.
Probst, D.; Reymond, J. L. A Probabilistic Molecular Fingerprint for Big Data Settings.
ChemRXiv 2018, doi.org/10.26434/chemrxiv.7176350.v1.
11.
Hagadone, T. R. Molecular Substructure Similarity Searching: Efficient Retrieval in Two-
Dimensional Structure Databases. J. Chem. Inf. Comput. Sci. 1992, 32, 515-521.

29
ACCEPTED MANUSCRIPT
Geppert, H.; Vogt, M.; Bajorath, J. Current Trends in Ligand-Based Virtual Screening:
12.
Molecular Representations, Data Mining Methods, New Application Areas, and Performance
Evaluation. J. Chem. Inf. Model. 2010, 50, 205-216.
13.
Owen, J. R.; Nabney, I. T.; Medina-Franco, J. L.; López-Vallejo, F. Visualization of
Molecular Fingerprints. J. Chem. Inf. Model. 2011, 51, 1552-1563.
14.
Scior, T.; Bender, A.; Tresadern, G.; Medina-Franco, J. L.; Martinez-Mayorga, K.; Langer,
T.; Cuanalo-Contreras, K.; Agrafiotis, D. K. Recognizing Pitfalls in Virtual Screening: A Critical
Review. J. Chem. Inf. Model. 2012, 52, 867-881.
15.
Riniker, S.; Landrum, G. A. Open-Source Platform to Benchmark Fingerprints for Ligand-
Based Virtual Screening. J. Cheminform. 2013, 5, 26.
16.
Rogers, D.; Hahn, M. Extended-Connectivity Fingerprints. J. Chem. Inf. Model. 2010, 50,
742-754.
17.
Mysinger, M. M.; Carchia, M.; Irwin, J. J.; Shoichet, B. K. Directory of Useful Decoys,
Enhanced (Dud-E): Better Ligands and Decoys for Better Benchmarking. J. Med. Chem. 2012, 55,
6582-6594.
18.
Nguyen, K. T.; Blum, L. C.; van Deursen, R.; Reymond, J.-L. Classification of Organic
Molecules by Molecular Quantum Numbers. ChemMedChem 2009, 4, 1803-1805.
19.
van Deursen, R.; Blum, L. C.; Reymond, J. L. A Searchable Map of Pubchem. J. Chem. Inf.
Model. 2010, 50, 1924-1934.
20.
Schwartz, J.; Awale, M.; Reymond, J.-L. Smifp (Smiles Fingerprint) Chemical Space for
Virtual Screening and Visualization of Large Databases of Organic Molecules. J. Chem. Inf. Model.
2013, 53, 1979-1989.
21.
Awale, M.; Reymond, J.-L. Atom Pair 2d-Fingerprints Perceive 3d-Molecular Shape and
Pharmacophores for Very Fast Virtual Screening of Zinc and Gdb-17. J. Chem. Inf. Model. 2014,
54, 1892-1907.
22.
Probst, D.; Reymond, J. L. Fun: A Framework for Interactive Visualizations of Large, High-
Dimensional Datasets on the Web. Bioinformatics 2018, 34, 1433-1435.