hazard assessment of mutagenic impurities
TRANSCRIPT
Sub Heading
Date
Hazard Assessment Of
Potentially Mutagenic Impurities
ICH M7 Indian Roadshows 2020
February 28, 2020
March 02, 2020
Muzaffar Khan, Ph.D
Introduction
2
• Hazard assessment of 8 example compounds at Laurus Labs
with varying expert rule-based and statistical-based results
• Derek Nexus and Sarah Nexus results will be presented
• A conclusion will be drawn for each compound based on
expert call
• You will be asked to assess each example and vote
accordingly
• Laurus/Lhasa will then present an expert call*
• Follow-up hazard assessment on the alerting compounds will
be discussed
*All expert calls on the examples are solely the decisions taken at Laurus Labs
and corroborated by Lhasa Limited.
Hazard Assessment - ICH M7
3
What kind of arguments can be used?
4
• Adequate Ames data available
• Ames test does not assess the hazard caused by the
compound class adequately
• Toxicophore identified by one system has not been
adequately assessed by the other
• Toxicophore identified is not causative of activity
• Nearest neighbours are not adequately similar to support
prediction
• Data available for nearest neighbours is not of sufficient
quality to support prediction
5
Example-1
6
Review high level predictions
7
Review the expert prediction
8
• No alerts fired
• Contains unclassified features
Review the expert prediction
9
• The test compound contains an unclassified feature (Imidoyl chloride)
which does not match with any structural alerts or examples for Ames
positive compounds in Derek.
• Predicted to be inactive in Ames mutagenicity test.
• Imidoyl chlorides are electrophilic in nature with a good leaving group
and therefore potentially mutagenic.
• Imidoyl chlorides are analogous of acyl chlorides and highly reactive.
• Imidoyl chlorides react spontaneously with water to yield the amides.
Review the statistical prediction
10
• Outside domain
• Training set does not include any example covering this functionality
Expert review
11
Inactive *
Outside domain
• Inactive prediction but contains unclassified features
• Potential alkylating agent (short lived as hydrolysis is highly likely)
• Out of domain
• Sarah Nexus training set does not contain examples covering
“imidoyl chloride”
Please make your selection
12
1. Class 3 –Alerting structure
2. Class 5 –No alerts or alerting with sufficient data to
demonstrate lack of mutagenicity
3. Unsure
M7 Classification – Example 1
13
• Derek Nexus
• Inactive prediction, contains unclassified features
• Potential alkylating agent; highly reactive
• Sarah Nexus
• Outside domain with no similar examples in the training set
M7 Classification – Example 1
14
Class 3 Alerting structure
Hazard Assessment & Control Strategy:
Impurity introduced too early in synthesis (n-6 step) and unlikely to carryover to
the drug substance. Proposed ICH M7 Option 4 control strategy based on
commercial batch data and spike-purge studies.
Mirabilis is useful to justify Option 4 control strategy
Example-2
15
CAS No. 173676-59-0
Review high level predictions
16
Review the expert prediction
17
• No alerts fired
• A certain aromatic amine and potentially mutagenic
• A more detailed review is necessary
Review the statistical prediction
18
• Positive hypothesis for aromatic amine• The most relevant parent (2-Acetylaniline) is non-mutagenic.
• The 4-chloroaniline containing compounds (2,3,4&5) are not relevant.
• A Vitic search indicates that 4-Chloroaniline has displayed activity in TA98 (albeit
not very consistently). Listed in ICH M7 R1 Addendum too.
Expert Review
19
• Inactive prediction with no misclassified or unclassified features
• Compound contains an aromatic amine
• Expert review of pAA is a subjective process and the in silico
results should be handled with care.
• The stability and physicochemical properties of a subset of
aromatic amines affords them capricious mutagenic properties
in the Ames test.
• Positive hypothesis for aromatic amine
• 2-Acetylaniline is non-mutagenic, no relevant examples in
training set for 4-Chloroaniline
• 4-Chloroaniline has displayed mutagenic activity in TA98.
Inactive
Positive
Please make your selection
20
1. Class 3 – Alerting structure
2. Class 5 – No alerts or alerting with sufficient data to
demonstrate lack of mutagenicity
3. Unsure
M7 Classification – Example 2
21
• Derek Nexus
• Inactive prediction with no misclassified or unclassified features.
• The 2-substituent (electron withdrawing group) deactivates the aromatic
amine, there is no suitable electron donating group at another position
• Disubstituted aniline bearing only halogen and/or trifluoromethyl
substituents on aromatic ring are generally Ames negative
• Sarah Nexus
• Positive prediction for aromatic amine
• The most relevant example is Ames negative.
M7 Classification – Example 2
22
Experimental results confirmed the compound was Ames negative
(Class 5)
Class 5 No alerts or alerting with sufficient data to demonstrate
lack of mutagenicity
Example-3
23
CAS No. 27607-77-8
Review high level predictions
24
Review the expert prediction
25
• No alerts fired
• Compound contains unclassified features
Review the statistical prediction
26
• Outside domain
• Nothing to report
Review the expert prediction
27
• Triflic acid predicted to be non-mutagenic
• No misclassified or unclassified features
Review the expert prediction
28
• TMS-OH predicted to be non-mutagenic
• No misclassified or unclassified features
Review the statistical prediction
29
• Negative prediction for triflic acid
Review the statistical prediction
30
• TMS-OH is Ames negative
Expert Review
31
• Inactive prediction
• Compound contains unclassified features
• Outside domain
• Nothing to report
Inactive *
Outside domain
Please make your selection
32
1. Class 3 –Alerting structure
2. Class 5 –No alerts or alerting with sufficient data to
demonstrate lack of mutagenicity
3. Unsure
M7 Classification – Example 3
33
• Derek Nexus
• Inactive prediction but contains unclassified features.
• No alerts fired for the potential hydrolysis products, triflic acid and TMS-OH
• Sarah Nexus
• Out of domain
• The hydrolysis product triflic acid is predicted negative
• TMS-OH is Ames negative
M7 Classification – Example 3
34
Class 5 No alerts or alerting with sufficient data to demonstrate
lack of mutagenicity
Example- 4
35
CAS NO:17933-03-8
Review high level predictions
36
Review the expert prediction
37
• Plausible mutagen
• Alert – 746: Arylboronic acid or derivative
Review the statistical prediction
38
Exact match -Positive Ames Call reported from Vitic Nexus
• 4 most relevant nearest neighbours are all positive
• Negative nearest neighbours are dissimilar.
Review the statistical prediction
Expert Review
40
• Plausible mutagen
• Alert – 746: Arylboronic acid or derivative
• No reason to over-rule Derek prediction
• Predicted positive -100% confidence (exact match)
• 4 most relevant nearest neighbours are all positive
• Ames data is reliable
Positive
Plausible
Please make your selection
41
1. Class 3 –Alerting structure
2. Class 5 –No alerts or alerting with sufficient data to
demonstrate lack of mutagenicity
3. Unsure
M7 Classification – Example 4
42
• Derek Nexus
• Plausible mutagen
• Alert – 746: Arylboronic acid or derivative
• Sarah Nexus
• Predicted positive -100% confidence (exact match)
• 4 most relevant nearest neighbours are all positive
• Negative nearest neighbours are not very relevant
M7 Classification – Example 4
43
Class 2 Mutagenic
Hazard Assessment & Control Strategy
Introduced in n-2 synthetic step; proposed Option-3 control strategy in
the starting material
Example- 5
44
CAS No. 25487-66-5
Review high level predictions
45
Review the expert prediction
46
• Predicted to be inactive
• Arylboronic acid or derivative?
Review the expert prediction
47
Aryl boronic acids - A mutagenic response is seen both with and
without metabolic activation, most commonly with Salmonella
typhimurium TA100 and TA1537 and Escherichia coli WP2 uvrA
strains. The mechanism by which these compounds exert their
mutagenic effect remains unclear.
Electron donating and sterically hindering groups are thought to
modulate reactivity of the boron centre, which may relate to mutagenic
activity. Strong correlation with 11B NMR shifts.
Following deactivating substituents on phenylboronic acids excluded.
- Carboxylic acid or tetrazole
- 2,6-Disubstitution
- Bulky alkyl para-substitution
- 3,4-Carbocyclic aliphatic fused ring substitution
Review the statistical prediction
48
• Equivocal
• Positive for Aryl boronic acid, however;
• Both the examples in training set are not very relevant as they do not
contain carboxylic acid moiety.
Expert Review
49
Arylboronic acid genotoxicity: Summary of data and knowledge
NMR shifts of arylboronic acids
Dr. Richard Williams, Lhasa Limitedhttps://www.lhasalimited.org/Public/Library/2015/Arylboronic%20acid%20genotoxicity%20-%20summary%20of%20data%20and%20knowledge.pdf
3-Carboxyphenylboronic acid was non-mutagenic in AMES assay
Stannard, L., Giddings, A., Lad, A., Munoz-Muriedas, J., Harvey, J., and Kenny, J. (2012)
Can lowest unoccupied molecular orbital (LUMO) energies predict the mutagenicity of
boronic acids? Mutagenesis 27, 807−808.
Expert Review
Expert Review
51
• Inactive prediction with no misclassified or unclassified features
• Compounds containing carboxylic acid substitution are
excluded from the alerts list
• Equivocal
• Positive hypothesis for the aryl boronic acid
• Nearest neighbours with positive results are not very relevant
• Training set does not contain phenylboronic acids with
carboxylic acid substitution
Inactive
Equivocal
Please make your selection
52
1. Class 3 –Alerting structure
2. Class 5 –No alerts or alerting with sufficient data to
demonstrate lack of mutagenicity
3. Unsure
M7 Classification – Example 5
53
• Derek Nexus
• Inactive prediction with no misclassified or unclassified features.
• No reason to over-rule Derek Nexus prediction
• Sarah Nexus
• Equivocal
• Positive examples not very relevant
• Compound is Ames negative (data available elsewhere)
M7 Classification – Example 5
54
Class 5 No alerts or alerting with sufficient data to demonstrate
lack of mutagenicity
Example- 6
55
Review high level predictions
56
Review the expert prediction
57
• Predicted to be inactive
• Contains unclassified features
Review the statistical prediction
58
• Positive hypothesis for Thiophene driven by presence of unrelated
toxicophores
Expert Review
59
• Inactive prediction, contains unclassified features
• Thiophene itself is not an alerting structure for mutagenicity
• Thiophene was non-mutagenic in Ames test*. *AESCHBACHER,HU, et.al CONTRIBUTION OF COFFEE AROMA CONSTITUENTS TO THE MUTAGENICITY OF COFFEE; FOOD CHEM. TOXICOL. 27(4):227-232, 1989.*ZEIGER,E, ANDERSON,B et.al. SALMONELLA MUTAGENICITY TESTS: III. RESULTS FROM THE TESTING OF 255 CHEMICALS; ENVIRON. MOL. MUTAGEN. 9(SUPPL.9):1-110, 1987.
• Aryl fused thiophene not necessarily Ames negative.
• Equivocal prediction
• All positive examples were confounded• All the Ames positive compounds in training set containing thiopene
fragment contained other known mutagenic functionalities (e.g.
aromatic nitro compounds).
Derek: Inactive *
Sarah: Equivocal
Please make your selection
60
1. Class 3 –Alerting structure
2. Class 5 –No alerts or alerting with sufficient data to
demonstrate lack of mutagenicity
3. Unsure
M7 Classification – Example 6
61
• Derek Nexus
• Inactive prediction with unclassified features.
• Sarah Nexus
• Equivocal
• Equivocal prediction driven by examples from training set containing
different toxicophores.
M7 Classification – Example 6
62
Class 5 No alerts or alerting with sufficient data to demonstrate
lack of mutagenicity
Example- 7
63
Review high level predictions
64
Review the expert prediction
65
• Positive prediction by Derek (Alert – 028: Mono- or di-alkylhydrazine)
• Hydrazines often produce weak and inconsistent responses in the Ames test
(due to their high toxicity and tendency to oxidise, which confounds testing)
and hence the compound is fired as an Equivocal alert in Derek.
Review the statistical prediction
66
• Predicted overall negative in Sarah Nexus
• Positive hypothesis for hydrazine
Expert Review
67
• Equivocal
• Alert – 028: Mono- or di-alkylhydrazine
• Alert as per ICH M7
• Negative prediction
• Positive hypothesis for alkyl hydrazine
• Compounds in the training set with same alerting functionality are
Ames positive
Equivocal
Negative
M7 Classification – Example 7
68
Imp-1
Imp-2Atazanavir API
Hazard Assessment
69
Structure Derek
Prediction
Sarah
Prediction
QSAR
Prediction
Experimental
Data
Similarity
to API
ICH M7
Class
Comments
Imp-1
Carc: Unspecified Ames: Unspecified
All Alerts found in API
EQUIVOCAL: Alert028 -
Mono- or di-alkylhydrazine
Class 4
Imp-2
Carc: Unspecified Ames: Unspecified
All Alerts found in API EQUIVOCAL: Alert028 -
Mono- or di-alkylhydrazine
Class 4
Please make your selection
70
1. Class 3 –Alerting structure
2. Class 4 –Alerting structure, same alert in drug substance
or compounds related to the drug substance (e.g.,
process intermediates) which have been tested and are
non-mutagenic
3. Class 5 –No alerts or alerting with sufficient data to
demonstrate lack of mutagenicity
Imp-1
M7 Classification – Example 7
71
• Derek Nexus
• Equivocal; Alert- 028: mono- or dialkyl - hydrazine
• Predicted as ICH M7 Class 4 but chemical environment of the alert
is essentially not same as that of the API
• Sarah Nexus
• Negative prediction
• Alkyl hydrazines in training set are Ames positive
M7 Classification – Example 7
72
Class 3 Alerting structure unrelated to the structure of API
Imp-1
M7 Classification – Example 7
73
Hazard Assessment & Control Strategy
Option-1: Control at TTC limit
Option-2: Perform Ames test to classify as Class 2 or 5
Imp-2: Ames positive (in Salmonella
tester strains TA1537 and TA98) !Imp-1 : Ames negative
Imp-2 was non-clastogenic in the follow-up in vivo micronucleus assay in rats.
Example- 8
74
Review high level predictions
75
Review the expert prediction
76
• Predicted to be inactive
• No misclassified or unclassified features
• Potential aromatic amine, aryl N-oxide and acrylonitrile present
Review the statistical prediction
77
• Positive hypothesis for aryl-N-oxide
• >=1 fused aromatic ring or second toxicophore is required for activity
Expert Review
78
• Inactive prediction, no misclassified or unclassified features
• Mutagenicity from the amine between two aromatic rings is unlikely
• Aryl-N-Oxide and Acrylonitrile alerts have not been fired for valid
SAR reasons
• Positive prediction
• Aromatic N-oxide alert
• All positive examples were confounded containing > 1 fused
aromatic rings or other alerts e.g. aromatic nitro or amine
Inactive
Positive
Expert Review – Public data
79
Myatt et al. concluded based on SARs that the general class of aromatic N-Oxide is not an alert for predicting DNA-reactive mutagenicity (except quindioxins)Mutagenesis, Volume 34, Issue 1, January 2019, Pages 67–82, https://doi.org/10.1093/mutage/gey020
Please make your selection
80
1. Class 3 –Alerting structure
2. Class 5 –No alerts or alerting with sufficient data to
demonstrate lack of mutagenicity
3. Unsure
M7 Classification – Example 8
81
• Derek Nexus
• Inactive prediction with no misclassified or unclassified features.
• No reason to over-rule Derek Nexus prediction
• Sarah Nexus
• Positive
• All positive examples are confounded, contain other alerts (e.g.
aromatic nitro or amine and > 1 aromatic rings) and not very
relevant
M7 Classification – Example 8
82
Class 5 No alerts or alerting with sufficient data to demonstrate
lack of mutagenicity
Conclusions
83
1. ICH M7 recommends using two complementary in-silico tools;
Expert rule-based (eg. Derek Nexus) and Statistical (eg. Sarah
Nexus).
2. Both expert and statistical tools are necessary!
3. Equivocal / out of domain results should be treated further with
expert opinion to arrive at meaningful conclusion.
4. Dependence only on literature based structural alerts (e.g.
Mueller’s alerts) for the expert review process, without
considering mitigating factors, could result in many more false
positives (Mayden A, Williams RV et al. (2017), Utility of published DNA
reactivity results, https://pubmed.ncbi.nlm.nih.gov/28549899.).
5. The use of Derek Nexus and Sarah Nexus significantly reduces
the number of false positives saving significant resources in the
hazard assessment process.
84
THANK YOU for your kind attention!
ACKNOWLEDGEMENTS
Dr. Richard Williams
LAURUS LABS LTD. and LHASA LIMITED