The use of Derek Nexus to facilitate decision-making in chemical safety assessmentCase studies

Principal Scientist

donna.macmillan@lhasalimited.org

Dr Donna Macmillan

Application of non-animal approaches for decision-making in chemical safety assessment10th-11th December, NC3Rs, London

Agenda

• Introduction to Lhasa Limited

• Derek Nexus

• Case studies:• Use as a screening tool in the pharmaceutical industry• Use in an ICH M7 regulatory context• Use in a defined approach for skin sensitisation

• Conclusions

Introduction to Lhasa Limited

Lhasa Limited

• Established 1983• Not-for-profit organisation • Educational charity• Headquarters in Leeds, UK• Offices in Newcastle, UK, Poznan, Poland• About 300 organisations worldwide use our software

• Pharmaceutical companies• Chemical and agrochemical companies• Personal product and cosmetic companies• Universities• Contract research organisations• Government (including regulatory) bodies• Consultants

Lhasa Limited - Products

Toxicity Prediction

Mutagenicity Prediction

Metabolism Prediction

Toxicity Database & Data Sharing

Degradation Prediction

Prediction of Impurity Purging

Derek Nexus

Derek Nexus

Case study 1: Use as a screening tool in the pharmaceutical industry

Derek Nexus in drug screening

• Widely used in initial drug development for toxicity (safety) profiling of target molecules• Between 20-30% of failures are generally attributed to safety

reasons

• Advantages include:• Fast• Not requiring synthesis of molecule• Can be applied at early stage drug development or to refine

candidate selection• Provides toxicity predictions for multiple endpoints

simultaneously

Brigo & Muster, Methods Mol Biol., 2016, 1425, 475-510

• In silico alerts followed up with in vitro (and maybe even in vivo) assays

• Local SAR models may be created in custom systems (Derek Knowledge Editor)

Derek Nexus in drug screening

Target identification

Lead identification

Lead optimisation

Candidate selection Phase 0

Drug development process

• Chemical libraries can be evaluated using Derek to provide a basic safety profile for each chemical

• Safety hazards identified but not necessarily used for decision-making at this early stage

• Relevant chemical scaffolds identified

• In silico analysis of target and off-target activities assessed

• Scaffolds with likely genetoxicity, carcinogenicity, hERGchannel blockade may be avoided at this stage

• Upon selection of a final candidate drug molecule, before moving into Phase 0 development, the main use of in silico tools is for ICH M7 compliance

Brigo & Muster, Methods Mol Biol., 2016, 1425, 475-510

Case study 2: Use in an ICH M7 regulatory context

• The use of two QSAR systems are permitted to predict the outcome of in vitro mutagenesis for a given impurity:• “The absence of structural alerts from two complementary

(Q)SAR methodologies (expert rule-based and statistical) is sufficient to conclude that the impurity is of no mutagenic concern”

• “If warranted, the outcome of any computer system-based analysis can be reviewed with the use of expert knowledge in order to provide… a rationale to support the final conclusion”

ICH M7 - Overview

Classification using 2 in silico modelsLikely to conclude positiveVery strong evidence would be needed to overturn both

predictions

UncertainLikely to conclude positive without strong evidence to

overturn a positive prediction

Likely to conclude positiveLack of a second prediction

suggests insufficient evidence to draw any other

conclusion

Prediction 1 (Derek)

Prediction 2(Sarah)

Positive

Positive

Positive

O.O.D. or equivocal

Positive

Negative

Negative

O.O.D. or equivocal

Negative

Negative

UncertainConservatively could assign as positive.

May conclude negative with strong evidence showing feature driving a ‘no prediction’ is

present in the same context in known negative examples (without deactivating features)

Likely to conclude negativeExpert review should support this conclusion – e.g. by assessing any

concerning features (misclassified, unclassified, potentially reactive...)

O.O.D. = out of domain

Barber et al. Reg. Tox. and Pharmacol. 2015, 73, 367

Expert review: Example

1-(4-hydroxyphenyl)prop-2-en-1-one95605-38-2

Review high level predictions

?

Expert Review

M7 classification

Expert review

?

Expert Review

M7 classification

• Plausible prediction – fires alert 664 for alpha, beta-unsaturated compounds

• Mutagenicity related to electronegativity and steric hindrance around the double bond

• Aryl substituents at C3 are typically negative or have reduced activity

• Negative prediction• Negative compounds mostly have an aryl substituent at

the C3 position• Most similar compound is negative, however there is no

protocol information

Expert summary

• Derek plausible prediction• No data to support overturning the prediction

• Sarah negative prediction• Many supporting examples are not relevant

?

Expert Review

M7 classification

M7 classification

?

Expert Review

M7 classification

Class 3Alerting structure

Case study 3: Use in a defined approach for skin sensitisation

The skin sensitisation AOP

Organism response

Organ response

Cellular response

20

Molecular Initiating Event Haptenation

T-cell activation

Activation of DC

Stress response

Skin sensitisation

Adverse Outcome Pathway

KE3

KE2

AO

MIE/KE1

KE4

Figure adapted from OECD 2012, The Adverse Outcome Pathway for Skin Sensitisation Initiated by Covalent Binding to Proteins Part 1: Scientific Evidence, Series on Testing and Assessment, No. 168.

Chemical structure& properties

Metabolism & penetration

Electrophilic substance

Assay

DPRA

LLNA

h-CLAT / U-SENS™ / IL-8 Luc

KeratinoSens™ / LuSens

GPMT (& Buehler) 406 (1992)

OECD TG

442D (2018)

429 (2010)

442C (2015)

442E (2018)

in vivo

in chemico / in vitro

HRIPT / HMT / patch test

MIE = Molecular Initiating EventKE = Key EventAO = Adverse Outcome

Lhasa’s hypothesis

21

• Apply exclusion criteria to chemicals based on known assay limitations and confidence in Derek predictions

• Ensures the most relevant information source(s) are used for a given chemical (class)

• Use Derek and assay(s) measuring the relevant KE until a concordant result is obtained - or a 2 out of 3 majority call

Pre-MIEKE4AO

DPRA

MIE

KeratinoSens™

LuSens

KE2 KE3

h-CLATU-SENS™IL-8 Luc

MIE = molecular initiating eventKE = Key EventAO = adverse outcome

Exclusion CriteriaExclusion Criteria Derek MIE KE2 KE3 Comment

Metabolism Prohapten ✓ ✗ ✓ ✓Assays lacking metabolic competency are

deprioritised as they are less likely to predict prohaptens well

logP> 3.5 ✓ ✓ ✓ ✗ Cell-based assays are deprioritised for

chemicals with a logP > 3.5 (KE3) and logP > 5 (KE2) as more lipophilic chemicals may

lack high solubility in these cell-based assays> 5 ✓ ✓ ✗ ✗

Lysine reactive Exclusive ✓ ✓ ✗ ✓

The Nrf2-ARE pathway is associated with cysteine binding - lysine-reactive chemicals

may not be reliably predicted

Reasoning level Equivocal ✗ N/A

Alerts with a likelihood of equivocal have less evidence of skin sensitisation potential than other likelihoods (e.g. certain) and are

thus deprioritised

Negative prediction

Misclassified features ✗ N/A Negative predictions with ‘misclassified

features’ or ‘unclassified features’ are deprioritised as these are associated with

higher uncertainty.Unclassified

features ✗ N/A

DPRA

KeratinoSens™LuSens

h-CLATU-SENS™

IL-8 Luc

Macmillan & Chilton, Reg. Tox. and Pharmacol., 2019, 101, 35

Defined approach decision tree

Potency prediction

modelNH2

NH2

1st assay

1st assay

1st assay

EquivocalNon-sensitiser with misclassified or unclassified features

2nd assay

3rd assay

2nd assay

2nd assay

2nd assay

CertainProbablePlausible

Non-sensitiserDoubted

ImprobableImpossible

Derek alert

outcomePotency category 5/6 (GHS no cat)

Query Use Derek outcome to determine decision tree branch

Prioritise in chemico/in vitro

assays using exclusion criteria

Potency category 1 (GHS 1A)Potency category 2 (GHS 1A)Potency category 3 (GHS 1B)Potency category 4 (GHS 1B)

Run in chemico/in vitro assays in order of AOP (MIE → KE2 → KE3) unless de-

prioritised by exclusion criteriaPotency prediction using

k- nearest neighbours modelHazard prediction

using ‘2 out of 3’ approach

Exclusion criteria

sensitiser

sensitiser

non-sensitiser

non-sensitiser

sensitiser

non-sensitiser

Blue italics = Derek outcomeRed arrow = positive resultGreen arrow = negative result

Macmillan & Chilton, Reg. Tox. and Pharmacol., 2019, 101, 35

Results - Hazard

BA = Balanced AccuracySe = SensitivitySp = Specificity

Macmillan & Chilton, Reg. Tox. and Pharmacol., 2019, 101, 35

Results - Potency (GHS)Defined approach prediction vs in vivo outcome

LLNAn = 174

Acc = 73%

Humann = 79

Acc = 76%

no cat no cat

no c

at

no c

at

Under-prediction

Over-prediction

Under-prediction

Over-prediction

Macmillan & Chilton, Reg. Tox. and Pharmacol., 2019, 101, 35

Conclusions

• Three case studies have shown how Derek Nexus can be used to inform decision making in chemical risk assessment.

• As a standalone tool for screening in early pharmaceuticaldiscovery

• Regulatory application where expert review of predictions isimportant (ICH M7)

• Within a defined approach, where results from in vitro assaysalso need to be considered

Acknowledgements

• Chris Barber

• Rich Williams

• Martyn Chilton