Accelerated Discovery of Novel Ponatinib Analogs with Improved Properties for the Treatment of Parkinson's Disease

Article abstract of DOI:10.1021/acsmedchemlett.9b00612

Parkinson's disease (PD) is a debilitating and common neurodegenerative disease. New insights implicating c-Abl activation as a driving force in PD have opened a new drug development avenue for PD treatment beyond the symptomatic relief by L-DOPA. BCR-Abl inhibitors, which include nilotinib and ponatinib, have been found to inhibit this process, and nilotinib has shown improvement in outcomes in a 12-patient, nonrandomized trial. However, nilotinib is a potent inhibitor of hERG, a cardiac K⁺ channel whose inhibition increases risk of sudden death. We used our machine learning approach to predict novel molecules that would inhibit c-Abl while also having minimal liability against hERG. Of our six novel compounds tested, we identified two that had c-Abl potencies comparable to nilotinib, but with significantly improved profiles regarding the hERG channel. Our best compound exhibited a hERG IC₅₀ of 12.1 μM (compared to nilotinib with an IC₅₀ of 0.45 μM and ponatinib with IC₅₀ of 0.767 μM). This work is a step forward for a machine learning enabled, multiparameter optimization of a chemical space and represents a significant advance in the development of novel Parkinson's therapies.

Full text of DOI:10.1021/acsmedchemlett.9b00612

Subscriber access provided by Uppsala universitetsbibliotek
Letter
Accelerated Discovery of Novel Ponatinib Analogs with
Improved Properties for the Treatment of Parkinson’s Disease
Thomas M Kaiser, Zackery W. Dentmon, Christopher E. Dalloul, Savita K Sharma, and Dennis C. Liotta
ACS Med. Chem. Lett., Just Accepted Manuscript • DOI: 10.1021/acsmedchemlett.9b00612 • Publication Date (Web): 12 Mar 2020
Downloaded from pubs.acs.org on March 13, 2020
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 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 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.
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 10
ACS Medicinal Chemistry Letters
1
2
3
4
5
6
7
Accelerated Discovery of Novel Ponatinib Analogs with
Improved Properties for the Treatment of Parkinson’s Disease
Thomas M. Kaiser*b, Zackery W. Dentmona, Christopher E. Dalloula, Savita K. Sharmaa and Dennis C. Liotta*a
a Department of Chemistry, Emory University, 1521 Dickey Dr., Atlanta, GA 30322
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
b
St Peter’s College, University of Oxford, New Inn Hall Street, Oxford, UK OX1 2DL
KEYWORDS: Parkinson’s disease, c-Abl inhibitors, hERG, QT prolongation, machine learning
ABSTRACT: Parkinson’s disease (PD) is a debilitating and common neurodegenerative disease. New insights
implicating c-Abl activation as a driving force in PD have opened a new drug development avenue for PD treatment
beyond the symptomatic relief by L-DOPA. BCR-Abl inhibitors, which include nilotinib and ponatinib, have been found
to inhibit this process and nilotinib has shown improvement in outcomes in a 12-patient, non-randomized trial. However,
nilotinib is a potent inhibitor of hERG, a cardiac K+ channel whose inhibition increases risk of sudden death. We used
our machine learning approach to predict novel molecules that would inhibit c-Abl while also having minimal liability
against hERG. Of our six novel compounds tested, we identified two that had c-Abl potencies comparable to nilotinib,
but with significantly improved profiles regarding the hERG channel. Our best compound exhibited a hERG IC50 of 12.1
M (compared to nilotinib with an IC50 of 0.45 M and ponatinib with IC50 of 0.767 M). This work is a step forward for a
machine learning enabled, multiparameter optimization of a chemical space and represents a significant advance in the
development of novel Parkinson’s therapies.
ACS Paragon Plus Environment

ACS Medicinal Chemistry Letters
Page 2 of 10
INTRODUCTION
1
2
3
4
5
6
7
8
9
Neurodegenerative disease, a category which includes Parkinson’s disease (PD) and Alzheimer’s disease, represents
a significant healthcare burden worldwide that is expected to increase as lifespans are prolonged.1-2 A 2015 United
Nations report on world population aging anticipates a doubling of the number of people aged 60 years or older over the
next 35 years, and this demographic is the most at risk for developing neurodegenerative disease.3 PD is the second
most common neurodegenerative disease, affecting 2-3% of those over 65 years of age.4 The classic motor symptoms
include bradykinesia, postural instability, resting tremor and muscle stiffness. Additionally, there is significant disability
for the patient in that dementia, depression, dysphagia, insomnia and loss of smell frequently develop as PD progresses.5
These non-motor symptoms are key determinants of the quality of life in Parkinson’s patients and are gaining recognition
as important aspects to consider in therapeutic development.5-6
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 underlying pathogenesis of PD is complex. A key component of the disease process is neurodegeneration due to
selective and progressive loss of dopaminergic neurons in the substantia nigra pars compacta.7 The intracellular
accumulation of α-synuclein and the subsequent formation of Lewy bodies in the brainstem and olfactory system is an
additional histopathological feature of PD, but neither Lewy body formation nor dopaminergic neuron loss of the
substantia nigra alone define Parkinson’s.8 Instead, it is the combination of these two histopathological features that
results in the disease.8-9
Current therapy for PD is entirely focused on symptomatic relief and there are severe side effects associated with most
of the first line agents in clinical use.10 For example, dyskinesias, exacerbation of insomnia, movement freezing and
dose failure are common consequences of L-DOPA/carbidopa therapy.11 However, new insights into the pathological
mechanisms have inspired a novel strategy for PD drug development. Work in both animal models of PD and in post-
mortem analysis of human tissue affected by Parkinson’s has revealed the role of oxidative stress in dopaminergic
neurons as a key pathogenic step in PD.2, 12 One of the key links between oxidative stress and neurodegeneration is c-
Abl, a kinase historically associated with the BCR-Abl fusion gene in certain leukemias.13 Recent work has linked
aberrations in c-Abl activity resulting from oxidative stress to neurodegeneration in both Alzheimer’s and Parkinson’s
diseases.14-18 c-Abl activation and its subsequent phosphorylation of parkin result in inhibition of the ubiquitin ligase
activity of parkin, leading to the accumulation of toxic parkin substrates, including parkin interacting substrate (PARIS)
and aminoacyl tRNA synthetase complex-interacting multifunctional protein 2 (AIMP2).19-20 The findings that both PARIS
and AIMP2 accumulate in familial PD caused by parkin mutations and in MPTP toxin mouse models of PD strongly
implicates c-Abl in the pathogenesis of the disease.19-20
Based on these findings, several groups have explored the use of the known BCR-Abl inhibitors that are used clinically
for treating chronic myeloid leukemia to also treat PD. This class of drugs, which includes, imatinib, nilotinib and ponatinib
among others (Figure 1), has exhibited activity against untranslocated c-Abl.21 Specifically, imatinib was found to restore
parkin ubiquitin ligase activity and reduce PARIS and AIMP2 levels in the MPTP mouse model.16-17 Similarly,
administration of nilotinib protected against MPTP-induced dopaminergic deficits and reversed the loss of dopamine
neurons, a result known to improve motor behavior.22-23 Additionally, a 12-patient, non-randomized phase I trial
conducted at Georgetown University demonstrated that nilotinib treatment improved motor and cognitive outcomes.24
A
follow-up study evaluating the pharmacokinetics and pharmacodynamics of nilotinib in PD patients found that a 200 mg
single dose of nilotinib was capable of increasing cerebrospinal fluid concentrations of dopamine metabolites.25
Currently, a phase II trial evaluating nilotinib in PD is underway (NCT02954978).
N
N
N
N
N
N
N
N
N
CF3
N
H
HN
N
N
NH
O
H
H
N
N
N
O
F3C
N
O
N
N
Imatinib
Nilotinib
Ponatinib
Figure 1. A selection of approved c-Abl inhibitors
There are, however, significant side effects associated with these drugs that may adversely impact their use in treating
Parkinson’s disease. Nilotinib is known to be a potent inhibitor of the human Ether-à-go-go-Related Gene (hERG), a
potassium channel that generates the repolarization current in the cardiac action potential. Inhibition of this repolarization
process leads to prolongation of the QT interval, which can progress to torsades de points and cardiac arrest. Nilotinib
ACS Paragon Plus Environment

Page 3 of 10
ACS Medicinal Chemistry Letters
thus carries a black box warning for QT prolongation and sudden death (2.1% of patients experience QT increases of
>60ms, while <1% experience QT increases of >500ms; sudden deaths reported in 0.6% of patients).26-27 In addition,
nilotinib is a potent inhibitor of CYP3A4, which could preclude its use in patients who are also taking drugs that are
metabolized by this enzyme. Similarly, ponatinib has a black box warning for vascular occlusion, heart failure and
hepatotoxicity.28-30
1
2
3
4
5
We were intrigued by the possibility of designing new chemical entities that exhibit both good activity against c-Abl and
minimal off-target liabilities against hERG. Given our previous successes in using machine learning to optimize the
properties of known drugs,34,35 we sought to design novel c-Abl inhibitors with these features.
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
RESULTS AND DISCUSSION
Our initial investigation began with the design of two machine learning workflows. The first employed a single Naïve
Bayes Network (NBN) exclusively designed to predict the potency of novel small molecules against c-Abl (the
compounds ranked from this workflow will be referred to as single-NBN compounds). The second used two NBNs
working in concert; one designed for potency against c-Abl and the other to predict molecules with IC50 values greater
than 1 M against hERG (compounds ranked with this method will be referred to as dual-NBN compounds). To this end,
we acquired and curated c-Abl and hERG datasets from ChEMBL and enhanced them with additional information from
the patent literature to train the desired NBNs.31-33 In addition, by prospectively testing both workflows, we could
experimentally determine if there was a potency reduction resulting from the additional filtering step in the dual-NBN
workflow. An overview of our machine learning approaches, as well as the general workflows employed, are briefly
discussed in our previous publications.34-35
In selecting the core scaffolds that would be evaluated by our algorithms, we considered the medicinal chemistry profile
of nilotinib, since it had shown disease modification in animal studies, as well as ponatinib, which has not yet been
evaluated in Parkinson’s disease. From our perspective, analogs of ponatinib provide two key advantages over nilotinib
analogs: the ponatinib scaffold is straightforward to prepare and the parent compound offers a significantly more potent
starting point for new analog creation. As a consequence, losses in c-Abl potency that may result from the inclusion of
a hERG NBN might still provide us with the opportunity to obtain low nanomolar inhibitors.
Workflows were constructed in Pipeline Pilot36 and used to determine the chemical space most amenable to address the
c-Abl / hERG selectivity problem. Using this approach, we were able to explore millions of synthetic candidates not
previously reported in the scientific or patent literature. The rankings of the top candidates produced by single-NBN (c-
Abl only) and dual-NBN (c-Abl / hERG) workflows are shown in Figure 2. The reader will note that several highly
prioritized candidates were not prepared because of the synthetic challenges they presented.
Single-NBN (cAbl only)
Rank
Dual-NBN (cAbl / hERG)
Rank
N
H
1st
1st
N
N
O
N
N
N
N
N
HN
O
OH
N
N
Cl
N
H
N
1
4
Cl
N
9th
3rd
N
N
N
O
N
N
HN
N
OH
N
Cl
N
N
N
N
H
H
O
2
5
N
11th
12th
N
N
O
N
N
N
HN
O
OH
O
N
N
H
Cl
N
N
6
3
Figure 2. Synthetic targets generated for the single-NBN and the dual-NBN workflows
ACS Paragon Plus Environment

ACS Medicinal Chemistry Letters
Page 4 of 10
1
2
3
4
5
6
7
8
Two synthetic approaches were used to prepare the candidate compounds, a linear synthesis for the single-NBN
compounds and a divergent synthesis for the dual-NBN series (Schemes 1 and 2). The linear synthesis began with
alkylation of the TBDPS and methyl ether piperazinyl alcohols (9a and 9b) by 7 or 8 to afford 10a, 10b and 11,
respectively (details are included in the accompanying Supporting Information). Suzuki coupling with cyclopropyl
boronic acid and subsequent hydrogenation produced 13a and 13b in >76% yield over two steps. Direct
hydrogenation of 11 afforded 14 in >95% yield. Compounds 13a, 13b and 14 were each allowed to react with the
indicated acyl chloride to afford amides 15-17 in 70%, 77% and 85% yield respectively. Sonogashira coupling of 17
with 18 afforded the target compound 3 in 31% yield, and the TBDPS protected 15 and 16 were similarly coupled
with 18 in 72% and 33% yield, respectively. Subsequent desilylation of these two compounds gave the target
compounds 1 and 2 in 91% and 73%, respectively.
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
OH
B
HO
H
N
Pd(OAc)2, P(Cy)3, K3PO4
5:1 PhMe:H2O
100 °C, wave, 6 h
Et3N
DCM, rt, 16 h
Pt/C, H2
EtOAc, rt, 20 h
Br
Br
N
N
+
N
N
O2N
N
71%
80%
>95% crude
N
N
Br
OR1
OR1
H2N
OR1
O2N
OR1
O2N
N
O2N
10a and 10b
7
12a and 12b
13a and 13b
R
1 = TBDPS
1 = Me
9a:
9b:
R
H
N
Et3N
DCM, rt, 16 h
Pt/C, H2
EtOAc, rt, 20 h
Cl
Cl
Cl
N
+
N
N
N
72%
>95% crude
O2N
N
OTBDPS
9a
OTBDPS
H2N
OTBDPS
Br
8
11
14
N
N
O
1)
I
N
18
Cl
X
X
N
H3C
PdCl2(MeCN)2, Ph3P, CuI,
DIPEA, DMF, rt, 16 h, 72%
O
N
DIPEA, DMAP
THF, rt, 16 h
N
O
N
N
I
N
OR1
13a, 13b and 14
N
OR1
N
H
H
77%
2) TBAF, THF, rt, 1 h, 73%
H3C
X = cPr; R1 = TBDPS
X = Cl; R1 =TBDPS
X = cPr; R1 = Me
H3C
X = cPr, R1 = H
X = Cl, R1 = H
X = cPr, R1 = Me
15:
1:
2:
3:
16:
17:
Scheme 1: Linear synthesis of single-NBN compounds (c-Abl alone)
The divergent synthesis for the dual-NBN compounds began with reduction and acylation of 19 to give the desired
amide 21 in greater than >77% yield over two steps (Scheme 2). Sonogashira coupling with alkyne 18 afforded key
intermediate 22 in 51% yield over two steps. From 22, target compounds 4 and 5 were accessed via Dess Martin
oxidation of the benzylic alcohol, followed by reductive amination with the requisite bicyclic amine in >30% yield over
three steps. Additionally, target compound 6 was prepared by chlorination of 22 and subsequent alkylation of the
resulting benzylic chloride with the indicated isoindoline followed by deprotection. This strategy was used for 6 as the
reductive amination conditions which were successful for 4 and 5 ultimately failed to provide 6. These compounds
were then evaluated for their biological activities against c-Abl, hERG and CYP3A4.
ACS Paragon Plus Environment

Page 5 of 10
ACS Medicinal Chemistry Letters
1
2
3
4
5
6
O
N
I
Cl
1)
18
N
N
H3C
Cl
Pt/C, H2
EtOAc, rt, 16 h
Cl
O
OTBDPS
DMAP, DIPEA
THF, rt, 16 h
PdCl2•(MeCN)2, PPh3, CuI,
DIPEA, DMF, rt, 16 h, 87%
I
N
Cl
O2N
H2N
20
H
OTBDPS
OTBDPS
> 90% crude
86%
H3C
2) TBAF, THF, rt, 16 h, 59%
21
19
7
8
9
H
N
1)
N
HN
n
H
N
N
N
HOAc, 4 Å mol sieves
DMSO, 50 °C, 2 h
Dess-Martin
DMSO, rt, 0.5 h
OH
O
N
O
O
O
N
NH
N
N
N
N
10
11
12
13
14
15
16
17
18
19
20
21
22
23
n
N
Cl
N
H
Cl
N
H
2) STAB, 50 °C, 1 h
30%
N
Cl
N
H
> 95% crude
H3C
H3C
H3C
22
23
4:
5:
n = 2
n = 1
OTBDPS
1)
HN
SOCl2,
N
N
0 °C - rt, 0.5 h
KI, DIPEA, DMF, rt, 16 h, 71%
2) TBAF, rt, 16 h, 41%
N
O
Cl
O
N
N
N
65%
OH
N
Cl
N
Cl
N
H
H
H3C
H3C
24
6
Scheme 2. Divergent synthesis of the dual-NBN compounds (cAbl / hERG)
Table 1 shows the biological activities of our compounds against the indicated targets. Compounds 1-3 are single-
NBN compounds evaluated solely by the c-Abl NBN and, as predicted, exhibited picomolar IC50’s against c-Abl.
However, like their parent structure, they also potently inhibited hERG, rendering them unsuitable for further clinical
development. Alternatively, compounds 4 and 5, designed with the dual-NBN, were highly active against c-Abl (<10
nM) and exhibited significantly improved IC50 values against hERG (i.e., 5.02 and 12.1 M, respectively). Additionally,
these compounds had IC50 values greater than 5 µM against CYP3A4. Moreover, compounds 4 and 5 were found to
have substantially greater human liver microsome half-lives than nilotinib, although they were not as good as
ponatinib (Table 1). Compounds 4 and 5 exhibit negligible toxicity in THLE-3 and HEK293 cell lines. Taken
collectively, these data indicate that 4 and 5 possess an improved profile relative to both nilotinib and ponatinib.
Importantly, while the submicromolar hERG inhibition of ponatinib makes it unsuitable for use in patient populations
that have cardiovascular issues, it’s unlikely that compounds 4 and 5 could ever achieve the double digit micromolar
in vivo concentrations required to significantly inhibit hERG.37
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
cAbl
IC50
(nM)
Mouse Liver
Microsomes
t1/2 (minutes)
Human Liver
Microsomes
t1/2 (minutes)
THLE-3 HEK293
CC50 CC50 (µM)
(µM)
hERG
IC50 (µM)
CYP3A4
IC50 (µM)
1
0.30
0.39
0.65
9.96
8.31
>30
0.34
0.10
0.08
5.02
12.1
0.78
0.45
0.77
11.7
13.4
13.5
5.58
7.28
4.61
0.58
11.4
-
-
-
-
-
2
-
-
-
3
-
-
-
-
4
10.95
16.78
26.45
7.89
31.94
26.87
39.84
157.53
7.89
182.4
17.0
14.5
5
> 50.0
> 50.0
6
-
-
-
-
-
-
Nilotinib
Ponatinib
5.56
0.31
TABLE 1. cAbl, hERG and CYP3A4 activities as well as metabolic stability and CC50 values for target compounds
ACS Paragon Plus Environment

ACS Medicinal Chemistry Letters
Page 6 of 10
1
2
3
4
5
6
7
8
Because of the high structural homology of the ATP-binding pockets of kinases, all c-Abl inhibitors run the risk of
inhibiting other kinases.38 However, inhibition of multiple kinases may not necessarily be deleterious, as Fowler et al.
have demonstrated that inhibition of the Discoidin Domain Receptors (DDR1/2), the platelet-derived growth factor
receptor α and β (PDGFRs) and the tyrosine kinase, Src, potentiate the effect of c-Abl inhibition.39 To evaluate the
kinase inhibition profile of 4 and 5, we had them screened at 10 micromolar concentrations against the panel of >300
kinases maintained by the University of Dundee.40-42 This screen revealed that at 10 µM nilotinib inhibited 9 kinases
at >50%. Similarly, ponatinib inhibited 73 kinases, while compounds 4 and 5 inhibited 31 and 43 kinases, respectively
(see Supporting Information). Table 2 shows the per cent activity remaining of selected kinases screened at 10 µM
against nilotinib, ponatinib, 4 and 5, as well as the IC50 values for 4 and 5. Compound 4 was very potent against
PDGFRα, modestly inhibited Src, but did not inhibit DDR2. Conversely, compound 5 inhibited PDGRFα, Src and
DDR2. Since our algorithm was not trained to predict kinase selectivity, we consider the inhibition of kinases known
to potentiate c-Abl by compounds 4 and 5 to be largely fortuitous. However, since the structure of each compound
encodes its intrinsic response in a given measurement, the structural similarities of our compounds to nilotinib and
ponatinib undoubtedly bias them to have somewhat comparable profiles. More importantly, compounds 4 and 5
represent an excellent starting point for designing next generation c-Abl inhibitors with more desirable properties.
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
Nilotinib
Ponatinib
4
5
kinase
ABL
% Activity Remaining
% Activity Remaining and [IC50 (µM)]
2
4
1
3
8 [0.0031]
70 [>20]
11 [0.0013]
14 [0.2862]
16 [0.0020]
15 [0.4313]
1 [0.0708]
5 [0.0195]
17 [0.0261]
2 [0.0846]
9 [5.3500]
13 [0.0448]
14 [0.2404]
DDR2
PDGFRa
Src
56
74
25
52
37
23
97
75
107
25
11
2
10 [0.0021]
23 [0.7097]
2 [0.0932]
6 [0.0505]
22 [0.0908]
6 [0.1381]
45 [3.3060]
14 [0.0336]
62 [>20]
EPHB2
EPHB3
Lck
1
29
4
EPHA4
NEK6
52
28
2
VEGFR
MAP4K3
Table 2. % Activity remaining and IC50 values of nilotinib, ponatinib, 4 and 5 for selected kinases
CONCLUSIONS
We utilized a machine learning paradigm to identify a new series of compounds that were simultaneously optimized
for on-target potency (c-Abl inhibition) and off-target safety (hERG). We employed a single NBN to predict novel,
highly potent ponatinib analogs. By using two NBNs in concert, one optimized for c-Abl potency and the other for
minimizing hERG liability, we identified novel compounds, 4 and 5, with single digit nanomolar potencies and only
modest inhibition of hERG. Although less potent than ponatinib, our best compounds compare favorably to nilotinib,
the current agent in clinical trials for Parkinson’s disease. Moreover, our best compounds exhibited a much improved
CYP3A4 profile, as well as improved metabolic stability as compared to nilotinib. Initial development revealed our
compounds to be non-toxic against THLE-3 cells and HEK293 cells. Also, compound 4 potently inhibited PDGFRα
and Src, while compound 5 inhibited PDGFRα, Src and DDR2, kinase profiles that recent data suggests will improve
efficacy in treating PD. These unexpected kinase profiles were not anticipated at the time of machine learning design
and serve to highlight the important role serendipity often plays in drug development. However, key biological
ACS Paragon Plus Environment

Page 7 of 10
1
2
3
4
5
6
7
ACS Medicinal Chemistry Letters
experiments remain that will demonstrate the utility of compounds 4 and 5. In particular, the cerebrospinal fluid
concentration that can be achieved by 4 and 5 will likely be a key predictor of their efficacy in treating PD, and this is
a key datapoint which will be acquired in future animal studies. Most importantly, however, by using machine
learning, we were able to significantly accelerate the hit-to-lead optimization of these kinase inhibitors, as well as
simultaneously address multiple defects in the properties of clinical agents. The work reported herein bodes wells for
the potential utility of the compounds designed in this study to treat Parkinson’s disease.
8
9
ASSOCIATED CONTENT
Supporting Information
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 Supplemental Information contains a complete description of the synthesis and characterization of all of the
compounds described in the manuscript. Complete details of the biological assessments is also included.
The Supporting Information (SI) is available free of charge on the ACS Publications website. The SI is provided as a
PDF file.
AUTHOR INFORMATION
Corresponding Authors
* Thomas Kaiser (for machine learning enquiries) and *Dennis Liotta (for medicinal chemistry enquiries)
ACKNOWLEDGMENT
The authors wish to thank Dr. Sameshnee Pelly for her excellent assistance in preparing this manuscript.
ABBREVIATIONS
PD - Parkinson’s Disease, L-DOPA - L-3,4-dihydroxy-phenylalanine, BCR - Breakpoint cluster region, PARIS - parkin
interacting substrate, AIMP2 - aminoacyl tRNA synthetase complex-interacting multifunctional protein 2 , NBN - Naïve
Bayes Network.
REFERENCES
1.
Elbaz, A.; Carcaillon, L.; Kab, S.; Moisan, F., Epidemiology of Parkinson’s disease. Revue Neurologique 2016, 172, 14-
16.
2.
3.
4.
Wyss-Coray, T., Ageing, neurodegeneration and brain rejuvenation. Nature 2016, 539, 180-186.
World Population Ageing 2015 (ST/ESA/SER.A/390). United Nations: New York, NY, 2015.
Poewe, W.; Seppi, K.; Tanner, C. M.; Halliday, G. M.; Brundin, P.; Volkmann, J.; Schrag, A.-E.; Lang, A. E., Parkinson
disease. Nature Reviews Disease Primers 2017, 3, 17013.
5.
6.
7.
8.
Chaudhuri, K. R.; Schapira, A. H. V., Non-motor symptoms of Parkinson’s disease: dopaminergic pathophysiology and
treatment. The Lancet Neurology 2009, 8, 464-474.
Parkinson's Disease: National Clinical Guideline for Diagnosis and Management in Primary and Secondary Care. Royal
College of Physicians: London, UK, 2006; Vol. 35.
Benarroch, E. E.; Cutsforth-Gregory, J. K.; Flemming, K. D., Mayo Clinic Medical Neurosciences: Organized by
Neurologic System and Level. 6th ed.; Oxford University Press: New York, NY, 2018.
Dickson, D. W.; Braak, H.; Duda, J. E.; Duyckaerts, C.; Gasser, T.; Halliday, G. M.; Hardy, J.; Leverenz, J. B.; Tredici, K.
D.; Wszolek, Z. K.; Litvan, I., Neuropathological assessment of Parkinson's disease: refining the diagnostic criteria. The
Lancet Neurology 2009, 8, 1150-1157.
9.
Braak, H.; Tredici, K. D.; Rüba, U.; Vos, R. A. I. d.; Steur, E. N. H. J.; Braak, E., Staging of brain pathology related to
sporadic Parkinson’s disease. Neurobiology of Aging 2003, 24, 197-211.
10. Poewe, W.; Mahlknecht, P.; Jankovic, J., Emerging therapies for Parkinson's disease. Current Opinion in Neurology 2012,
25, 448-459.
11. Jankovic, J.; Poewe, W., Therapies in Parkinson's disease. Current Opinion in Neurology 2012, 25, 433-447.
12. Zhang, Y.; Dawson, V. L.; Dawson, T. M., Oxidative Stress and Genetics in the Pathogenesis of Parkinson's Disease.
Neurobiology of Disease 2000, 7, 240-250.
13. Druker, B. J.; Lydon, N. B., Lessons learned from the development of an Abl tyrosine kinase inhibitor for chronic
myelogenous leukemia. The Journal of Clinical Investigation 2000, 105, 3-7.
14. Li, B., c-Abl in oxidative stress, aging and cancer. Cell Cycle 2005, 4, 201-203.
15. Schlatterer, S. D.; Acker, C. M.; Davies, P., c-Abl in Neurodegenerative Disease. Journal of Molecular Neuroscience 2011,
45, 445-452.
ACS Paragon Plus Environment

ACS Medicinal Chemistry Letters
Page 8 of 10
1
2
3
4
5
6
7
8
16. Imam, S. Z.; Zhou, Q.; Yamamoto, A.; Valente, A. J.; Ali, S. F.; Bains, M.; Roberts, J. L.; Kahle, P. J.; Clark, R. A.; Li, S.,
Novel Regulation of Parkin Function through c-Abl-Mediated Tyrosine Phosphorylation: Implications for Parkinson's
Disease. The Journal of Neuroscience 2011, 31, 157-163.
17. Ko, H. S.; Lee, Y.; Shin, J.-H.; Karuppagounder, S. S.; Gadad, B. S.; Koleske, A. J.; Pletnikova, O.; Troncoso, J. C.;
Dawson, V. L.; Dawson, T. M., Phosphorylation by the c-Abl protein tyrosine kinase inhibits parkin's ubiquitination and
protective function. Proceedings of the National Academy of Sciences 2010, 107, 16691-16696.
18. Mahul-Mellier, A.-L.; Fauvet, B.; Gysbers, A.; Dikiy, I.; Oueslati, A.; Georgeon, S.; Lamontanara, A. J.; Bisquertt, A.;
Eliezer, D.; Masliah, E.; Halliday, G.; Hantsche, O.; Lashuel, H. A., c-Abl phosphorylates α-synuclein and regulates its
degradation: implication for α-synuclein clearance and contribution to the pathogenesis of Parkinson's disease. Human
Molecular Genetics 2014, 23, 2858-2879.
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
19. Shin, J.-H.; Ko, H. S.; Kang, H.; Lee, Y.; Lee, Y.-I.; Pletinkova, O.; Troconso, J. C.; Dawson, V. L.; Dawson, T. M., PARIS
(ZNF746) Repression of PGC-1α Contributes to Neurodegeneration in Parkinson's Disease. Cell 2011, 144, 689-702.
20. Lee, Y.; Karuppagounder, S. S.; Shin, J.-H.; Lee, Y.-I.; Ko, H. S.; Swing, D.; Jiang, H.; Kang, S.-U.; Lee, B. D.; Kang, H.
C.; Kim, D.; Tessarollo, L.; Dawson, V. L.; Dawson, T. M., Parthanatos mediates AIMP2-activated age-dependent
dopaminergic neuronal loss. Nature Neuroscience 2013, 16, 1392-1400.
21. Rossari, F.; Minutolo, F.; Orciuolo, E., Past, present, and future of Bcr-Abl inhibitors: from chemical development to
clinical efficacy. Journal of Hematology & Oncology 2018, 11, 84.
22. Karuppagounder, S. S.; Brahmachari, S.; Lee, Y.; Dawson, V. L.; Dawson, T. M.; Ko, H. S., The c-Abl inhibitor, Nilotinib,
protects dopaminergic neurons in a preclinical animal model of Parkinson’s disease. Scientific Reports 2014, 4, 4874.
23. Hebron, M. L.; Lonskaya, I.; Moussa, C. E.-H., Nilotinib reverses loss of dopamine neurons and improves motor behavior
via autophagic degradation of α-synuclein in Parkinson's disease models. Human Molecular Genetics 2013, 22, 3315-3328.
24. Pagan, F.; Hebron, M.; Valadez, E. H.; Torres-Yaghi, Y.; Huang, X.; Mills, R. R.; Wilmarth, B. M.; Howard, H.; Dunn,
C.; Carlson, A.; Lawler, A.; Rogers, S. L.; Falconer, R. A.; Ahn, J.; Li, Z.; Moussa, C., Nilotinib Effects in Parkinson’s
Disease and Dementia with Lewy Bodies. Journal of Parkinson's Disease 2016, 6, 503-517.
25. Pagan, F. L.; Hebron, M. L.; Wilmarth, B.; Torres-Yaghi, Y.; Lawler, A.; Mundel, E. E.; Yusuf, N.; Starr, N. J.; Arellano,
J.; Howard, H. H.; Peyton, M.; Matar, S.; Liu, X.; Fowler, A. J.; Schwartz, S. L.; Ahn, J.; Moussa, C., Pharmacokinetics
and pharmacodynamics of a single dose Nilotinib in individuals with Parkinson's disease. Pharmacology Research &
Perspectives 2019, e00470.
26. Doherty, K. R.; Wappel, R. L.; Talbert, D. R.; Trusk, P. B.; Moran, D. M.; Kramer, J. W.; Brown, A. M.; Shell, S. A.;
Bacus, S., Multi-parameter in vitro toxicity testing of crizotinib, sunitinib, erlotinib, and nilotinib in human cardiomyocytes.
Toxicology and Applied Pharmacology 2013, 272, 245-255.
27. Goldenberg, M. M., Pharmaceutical Approval Update. Pharmacy and Therapeutics 2008, 33, 54-57.
28. Rivera, V. M.; Pritchard, J. R.; Gonzalvez, F.; Baker, T.; Gozgit, J. M.; Hodgson, G., Comparative TKI Profiling Analyses
to Explore Potential Mechanisms of Ponatinib-Associated Arterial Thrombotic Events. Blood 2014, 124, 1783.
29. Latifi, Y.; Moccetti, F.; Wu, M.; Xie, A.; Packwood, W.; Qi, Y.; Ozawa, K.; Shentu, W.; Brown, E.; Shirai, T.; McCarty,
O. J.; Ruggeri, Z.; Moslehi, J.; Chen, J.; Druker, B. J.; López, J. A.; Lindner, J. R., Thrombotic microangipathy as a cause
for cardiovascular toxicity from the BCR-ABL1 tyrosine kinase inhibitor ponatinib. Blood 2019, 133, 1597-1606.
30. Hoy, S. M., Ponatinib: a review of its use in adults with chronic myeloid leukaemia or Philadelphia chromosome-positive
acute lymphoblastic leukaemia. Drugs 2014, 74, 793-806.
31. P., B. A.; Gaulton, A.; Hersey, A.; Bellis, L. J.; Chambers, J.; Davies, M.; Kruger, F. A.; Light, Y.; Mak, L.; McGlinchey,
S.; Nowotka, M.; Papadatos, G.; Santos, R.; Overington, J. P., The ChEMBL bioactivity database: an update. Nucleic Acids
Research 2014, 42, 1083-1090.
32. Papadatos, G.; Gaulton, A.; Hersey, A.; Overington, J. P., Activity, assay and target data curation and quality in the
ChEMBL database. Journal of Computer Aided Molecular Design 2015, 29, 885-896.
33. Gaulton, A.; Bellis, L. J.; Bento, A. P.; Chambers, J.; Davies, M.; Hersey, A.; Light, Y.; McGlinchey, S.; Michalovich, D.;
Al-Lazikani, B.; Overington, J. P., ChEMBL: a large-scale bioactivity database for drug discovery. Nucleic Acids Research
2012, 40 (D1), D1100-D1107.
34. Kaiser, T. M.; Burger, P. B.; Butch, C. J.; Pelly, S. C.; Liotta, D. C., A Machine Learning Approach for Predicting HIV
Reverse Transcriptase Mutation Susceptibility of Biologically Active Compounds. Journal of Chemical Information and
Modeling 2018, 58, 1544-1552.
35. Shi, Q.; Kaiser, T. M.; Dentmon, Z. W.; Ceruso, M.; Vullo, D.; Supuran, C. T.; Snyder, J. P., Design and Validation of
FRESH, A drug Discovery Paradigm Resting on Robust Chemical Synthesis. ACS Medicinal Chemistry Letters 2015, 6,
518-522.
36. BIOVIA Pipeline Pilot. https://www.3dsbiovia.com/products/collaborative-science/biovia-pipeline-pilot/ (accessed
February 16, 2020)
37. Compound 6 failed to meet our targeted profile (c-Abl IC50 value >30 nM; hERG IC50 = 780 nM) indicating that our
algorithms need further refinement. .
38. Robinson, D. D.; Sherman, W.; Farid, R., Understanding Kinase Selectivity Through Energetic Analysis of Binding Site
Waters. ChemMedChem 2010, 5, 618-627.
ACS Paragon Plus Environment

Page 9 of 10
1
2
3
ACS Medicinal Chemistry Letters
39. Fowler, A. J.; Hebron, M.; Missner, A. A.; Wang, R.; Gao, X.; Kurd‑Misto, B. T.; Liu, X.; Moussa, C. E. H., Multikinase
Abl/DDR/Src Inhibition Produces Optimal Effects for Tyrosine Kinase Inhibition in Neurodegeneration. Drugs in R&D
2019, 19, 149-166.
4
5
6
7
8
9
40. Kitagawa, D.; Yokota, K.; Gouda, M.; Narumi, Y.; Ohmoto, H.; Nishiwaki, E.; Akita, K.; Kirii, Y., Activity‐based kinase
profiling of approved tyrosine kinase inhibitors. Genes to Cells 2013, 18, 110-122.
41. Hastie, C. J.; McLauchlan, H. J.; Cohen, P., Assay of protein kinases using radiolabeled ATP: a protocol. Nat Protoc 2006,
1 (2), 968-71.
42. Bain, J.; Plater, L.; Elliott, M.; Shpiro, N.; Hastie, C. J.; McLauchlan, H.; Klevernic, I.; Arthur, J. S.; Alessi, D. R.; Cohen,
P., The selectivity of protein kinase inhibitors: a further update. Biochem J 2007, 408 (3), 297-315.
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

ACS Medicinal Chemistry Letters
Page 10 of 10
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
ACS Paragon Plus Environment