An in silico tool to aid identification of degradants from ...
37
An in silico tool to aid identification of degradants from forced degradation studies Senior Research Scientist [email protected] Dr Rachel Hemingway
Transcript of An in silico tool to aid identification of degradants from ...
An in silico tool to aid identification of degradants from forced degradation studies
Senior Research Scientist
Dr Rachel Hemingway
Aim of the talk
➢ To provide an overview of the current Zeneth
program
➢ To outline how the knowledge base is developed
➢ To detail the new scientific features being added
into the re-developed software
Contents
• Introduction to Lhasa Limited
• The Zeneth program
• Developing the knowledge base in Zeneth
• New scientific developments
• Future aspirations
• Concluding remarks
INTRODUCTION
Lhasa Limited
➢ Established 1983, headquarters in Leeds, UK
➢ Not-for-profit organisation and educational charity
▪ “Shared Knowledge, Shared Progress”
➢ It is controlled by its members (more than 350 organisations worldwide)
▪ Pharmaceutical companies▪ Chemical and agrochemical companies▪ Personal product and cosmetic companies▪ Food/Flavour companies▪ Tobacco companies▪ Universities▪ Contract research organisations▪ Government (including regulatory) bodies▪ Consultants
Lhasa: about us
Expert system
Database
Purge Factor
calculation tool
Statistical/
machine
learning
Vitic
Toxicity
Sarah Nexus
Mutagenicity Mirabilis
Purge Factors
Derek Nexus
Toxicity
Meteor Nexus
Xenobiotic metabolism
Zeneth
Chemical degradation
https://www.lhasalimited.org
THE ZENETH PROGRAM
Degradation workflow
Forced Degradation
Stability-Indicating Method
Stability Modelling
Final Formulation
Formulation Development
Packaging
Degradant Identification
Degradation Mechanism
In silico approach
Forced Degradation
Stability-Indicating
Method
Stability Modelling
Final Formulation
Formulation Development
Packaging
Degradant Identification
Degradation Mechanism
API
In silico assumption
RealSpace
PotentialSpace
HypotheticalSpace
A
B
C
D
E
F
GH
I
API
Hypothetical degradation products as in stress testing conditions and in silico
methods
BC
D
E
H
API
Potential degradation products as in accelerated and stability studies
BC H
API
Real degradation products in final packaging and storage conditions
Zeneth: an in silico tool
Z is for……….
• Expert knowledge based system for the prediction of forced degradation1, 2
• The program uses its knowledge base to:➢ find ‘matching’ transformations,
➢ predict chemical degradation,
➢ assess the likelihood of a given degradant
Transformation
Knowledge base
• Zeneth contains a ‘library’ of chemical transformations
1. An Expert System To Predict the Forced Degradation of Organic Molecules, Parenty DC, Button WG, Ott MA, Mol.
Pharmaceutics, 2013, 10, 292-2974, 10.1021/mp400083h.
2. Chapter 3: In silico drug degradation prediction, Ali MA, Hemingway R, Ott MA, in: Methods for Stability Testing of
Pharmaceuticals. Editors: Sanjay Bajaj, Saranjit Singh, pp 53-73.
Chemical patterns
Chemical degradation reactions are represented as patterns in the knowledge base
1. Degradation reaction
2. Determine the ‘degradophore’
3. Define the scope of the transformation, masking any proprietary nature
Excipients
• Potential API-excipient interactions can be predicted by Zeneth
• Important use case for many members
• Database of 188 structures (excipients, counterions and their associated
degradants and impurities)
Esterification with primary
aliphatic alcohol
Methodology
Query
Likelihood Threshold
Knowledge base Transformation Library
Environmental conditions
Reasoning
Degradant
Relative
Compares the likelihood of competing transformations
Absolute
Evaluates the likelihood that a degradation will
occur
Inp
ut
Tran
sfo
rmat
ion
En
gin
eAPI (+ excipients)
Environmental conditions
*Information from Q3B guideline
Methodology
Query
Likelihood Threshold
Knowledge base Transformation Library
Environmental conditions
Reasoning
Degradant
Relative
Compares the likelihood of competing transformations
Absolute
Evaluates the likelihood that a degradation will
occur
Inp
ut
Tran
sfo
rmat
ion
En
gin
eAPI (+ excipients)
Worked example
Query
Likelihood Threshold
Knowledge base Transformation Library
Environmental conditions
Reasoning
Degradant
Relative Absolute
Inp
ut
Tran
sfo
rmat
ion
En
gin
e
Conditions:Water, pH 4
Very Likely
N
R2
R1
R3
O
R2
R1 NH2
R3
H2O+
R1, R2 = Carbon or HydrogenR3 = Aliphatic or aromatic carbonC=N bond cannot be in a ring
Transformation 009
DEVELOPING THE KNOWLEDGE BASE
Knowledge base status
Knowledge base growth
Year No. of
transformations
Increase
2013 300 -
2014 367 +67
2015 400 +33
2016 446 +46
2017 492 +46
Knowledge base Transformation Library
Oxidation (152)
Hydrolysis(110)
Condensation (93)
Elimination (59)
Isomerisation(36)
Photolysis(42)
Total = 492
Industry influence
➢ Run brainstorming sessions with a group of Zeneth users
➢ We put forward proposals for inclusion in the knowledge base
➢ Industry can (and do) suggest proposals
➢ These decisions directly influence the chemistry that goes into the
knowledge base
Pfizer
Industry proposals
Norrish-Yang cyclisationBoron chemistry
+ hydrolysis of borinic, boronic and boric esters
Knowledge base Transformation LibraryRecent additions to the
NEW SCIENTIFIC DEVELOPMENTS FOR ZENETH 8
Re-development
What’s new?
Degradants will
be ‘scored’
instead of being
given a defined
likelihood level
Inclusion of a
property
calculator to
calculate pKa
values
Improvements in
treatment of
stereochemistry
Ability to handle
trimolecular
reactions
Scoring methodology
Moving from likelihood levels to scores
lead to more granular predictionsVery likely
Degradant A
0 2 4 6 8 10 12 14 16
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Formula = MIN((pH-1)/10 +0.1, 0.9)
Unlikely
Very unlikely
Equivocal
Likely
Very likely
Certain
Increasing
likelihood level
Impossible
0.90**
**scale used for illustrative purposes only, subject to change
pH profile 10 (base catalysed)
1 3 5 7 9 11 13pH
Very likely
Likely
Equivocal
Unlikely
Very unlikely
pH = 4 Equivocal
pH = 5 Equivocal
pH = 5.5 Equivocal
pH = 4 0.40
pH = 5 0.50
pH = 5.5 0.55
What are my top 20
most likely
degradants?
pKa calculator
• pKa of a molecule determines its protonation state at a given pH
• pKa tends to be governed by electronic effects that decrease with the distance from
the centre of protonation
• Generate a distance spectrum for each atom-type from the pKa centre
Dist 101101 101108 106307 106203 108103 108106 117100
1 1 1
2 0.25 0.25
3 0.22 0.11
4 0.12 0.06
Sum 1 0.34 1 0.36 0 0.25 0.06
Sybyl atom type
= distance spectrum
pKa performance
RMSE: 1.163
Preliminary results using 5-fold cross-validation
How will this calculator improve predictions?
pH = basic
Stereochemistry
• Have to define stereochemistry to garner stereochemical information
in any resulting degradants➢ No longer get individual isomers as degradants from an epimerisation reaction
unless the query had its stereochemistry defined
• SN2-type reactions will now demonstrate inversion of configuration
Transformation 134: Nucleophilic ring opening of aziridine
Trimolecular reactions
The new platform gives us the capabilities to express trimolecular reactions
FUTURE ASPIRATIONS FOR ZENETH
Other calculators
• Calculating BDEs is not trivial
• Balance between accuracy and speed of performance is a critical factor
• QM based calculation……………..a cheap one!
BDE calculator
Chromophore calculator
• Assess (the important) wavelength
• Applied as a whole-molecule property
• Modulate the prediction of photochemical reactions
Reducing false positives
• Zeneth predicts in the hypothetical chemical space
• There will always be a level of overprediction in comparison to the potential (or real) space
• Assessing how to reduce the number of false positives using a validation study
(Small) dataset of ~30 compounds with experimental data
(forced deg, accelerated and stability studies)
Are there
transformations which
fire for several APIs
which are never
actually seen?
➢ Scope amendment?
➢ Lower the score given?
What are we not
predicting that is
seen?
➢ Knowledge gaps
➢ Expand knowledge base
vs Hypothetical SpaceReal
SpacePotential
Space
Ideal in silico scenario
BC F
API
Real degradation products in final packaging and storage conditions
BC
D
E
F
API
Potential degradation products as in accelerated and stability studies
A
B
C
D
EF
API
Hypothetical degradation products as in stress testing conditions and in silico
methods
RealSpace
Hypothetical Space
PotentialSpace
RealSpace
PotentialSpace
Hypothetical Space
Concluding remarks
Zeneth is our
expert knowledge
based system for
the prediction of
forced
degradation
Helps understand
mechanistic
pathways as well
as aid degradant
identification
Zeneth generates
predictions from
its knowledge
base which
currently contains
492
transformations
Industry helps
drive the
development of
the knowledge
base and directly
influences the
content
Scientific
developments include
a granular scoring
method, improved
stereochemical
behaviour and
inclusion of a
calculator for pKa
Zeneth is
undergoing a
complete re-design
with the aim of
improving usability
as well as the quality
and speed of
predictions
Come see us!
Booth number: 10
Product Owner
for Zeneth
Lisa Leivers Maggie Coombs
US Account
Manager for Zeneth
Alison Reeves
Sales executive
Acknowledgements
Thank you all for your attention
Dr Martin Ott
Principal ScientistDr Carlos Cunha
(ex)-US Account Manager
Dr Jeff Plante
Senior Research Cheminformatician
Special thanks to:
Questions?
