14

Presentation 3

Applying mathematical modelling to crop protection development

Dr Kim Travis; Syngenta

By the late 1980s, mathematical modelling was used internally by the computational chemistry and chemical

engineering groups, and had recently started to be used by regulatory authorities in the USA, but experimental

approaches dominated. Whilst the discovery and development of a new crop protection product is still

dominated by empirical approaches, the use of mathematical modelling for internal decision-making has

considerably increased. The use of modelling to support regulatory decision-making has changed relatively

little in the USA, but in the EU has been totally transformed. Loud voices calling for model validation have

gradually been replaced by the recognition of the advantages of a model-based approach in the regulatory

process. In some domains the process has even gone too far, such that data is not believed if seems to be out

of line with the models! There has been most progress in modelling of the environmental fate and risk

assessment of chemicals, whilst change in mammalian toxicology and food residue assessment has been

much slower. I will try to identify the reasons why the acceptance and adoption of mathematical modelling has

varied so much in different scientific domains, internally vs externally, and between different regions.

Biography

Kim Travis is a biologist by training, who has a long standing interest in the modelling of biological systems. He

was employed in a Syngenta legacy company as a mathematical modeller in 1988. Since then he has worked

on all aspects of the safety of crop protection products, and has worked on projects ranging from new

chemical discovery through to supporting products that have been on the market for over 50 years. Kim is

currently a Syngenta Fellow, based at the Jealott’s Hill site in Berkshire. Syngenta is the world’s largest crop

protection company.

Kim Travis Sept 2016

Applying mathematical modelling to crop protection development

2

Outline

● Introduction● Uses of modelling within Syngenta● Uses of modelling in pesticide regulation● What enables the use of modelling in regulation?● Opportunities

3

Crop protection

● This means herbicides, fungicides and insecticides, i.e. pesticides- Protect yields, keep food affordable and meet population growth food

security challenge● Much as for traditional pharmaceuticals

- Small biologically-active molecules- Large discovery and screening effort- Many years and a huge amount of money from discovery to sales- Very strict regulatory process

● Unlike pharmaceuticals- We do far more ecological/environmental safety work- We don’t generally test our molecules in humans

4

Why are models used? Because they are useful!

Mathematical models express our ideas of how the world works in terms of equations

5

Modelling can save time, reduce risk and save money

…and deliver 3Rs benefits!

6

Modelling is everywhere in Syngenta

Model organisms Model environments

7

Animal studies and mathematical models are all models- we shouldn’t be thinking of them as fundamentally different

The issue with modelling is often not so much not validation, but acceptance

8

Outline

● Introduction● Uses of modelling within Syngenta

● Uses of modelling in pesticide regulation● What enables the use of modelling in regulation?● Opportunities

9

Experimentation Modelling

Insight

Foresight

virtuous circle

better understanding

prediction

10

Predictive toxicology

Existing knowledge in silico in vitro in vivo

11

• Acute toxicity (rats)• General repeat dose toxicity (rats, mice, dogs)• Genetic toxicity (in vitro, in vivo in rats and mice)• Carcinogenicity (rats, mice)• Development toxicity (rats, rabbits)• Reproductive toxicity over generations (rats)• Absorption, Distribution, Metabolism & Excretion (ADME)

Pesticide regulatory toxicology studies

12

Outline

● Introduction● Uses of modelling within Syngenta● Uses of modelling in pesticide regulation

● What enables the use of modelling in regulation?● Opportunities

13

Pesticide regulation

● Is it safe?- convince yourself- convince regulators too!

● Regulatory approval is a pre-requisite for product sales

14

Safety to manSafety to the

environment

Mammalian

toxicologyChemical fate

and exposure Ecotoxicology

The use of models in pesticide regulation

Niche uses Dominant role Increasing use

15

Role of environmental fate modelling in pesticide regulation in the EU

● Major role envisaged when EU rules were drafted in early ‘90s- For a new pesticide, environmental contamination could take years

to appear, so only models could address the need to ensure it could be protected against

- Then had to work out how to achieve the regulatory objectives with modelling FOCUS

• a forum for regulators, government institutes and industry to work out the details (mid ‘90s to mid ’00s)

16

Rain

Soil Properties

Leaching

Runoff

EvaporationTranspiration

17

18

19

EU pesticide leaching assessment- the original FOCUS standard scenarios

20

Standardised• Weather• Soil• Cropping

So regulatory evaluation can focus on pesticide properties• Use pattern• Soil absorption• Metabolism• ……

21

- maximises value of expensive studies- fits studies into a risk assessment- put study results into a real-world context

Use of models to aid study interpretation

22

Outline

● Introduction● Uses of modelling within Syngenta● Uses of modelling in pesticide regulation● What enables the use of modelling in regulation?

● Opportunities

23

What have we learnt?

● Mathematical modelling using standard scenarios can provide a level playing field for chemical regulation

But you also need● Good version control● A home (“institutionalisation”)● Documentation

…Models of ecological population dynamics- because the regulatory goal is to protect populations, not individuals- in crop protection, led to a large degree by my Syngenta colleague

Pernille Thorbek

24

● “…the current practice of modelling as published in the scientific literature needs some adjustments before it can be applied to

regulatory risk assessments.”● “To achieve this, modellers will have to follow good modelling practice.

A good start would be to include comprehensive and coherent

model descriptions, as many of the models that were reviewed did not include a full description of model design and input parameters.

25 PAGE 25

What do we need to base decisions on ecological models?

● CREAM project http://cream-itn.eu/

26 PAGE 26

THE MODELING CYCLE

Schmolke A, Thorbek P, DeAngelis DL, Grimm V. 2010.

Trends in Ecology and Evolution 25: 479-486

27

Verification and sensitivity analysis

● Verification: checking implementation- does the model do what you think it does?

● Sensitivity analysis- what parameters have the greatest influence on the outputs?- are there interactions between parameters?

● For complex models also model exploration: why do the model produce the patterns it does?

28

It’s not just about what the modeller needs to do…

Stuff model “customer” needs to do

Stuff to discuss with modeller

Stuff modeller should do

Stuff to ask an independent expert about

Stuff model assessor should do

29

Model communication

● Important for acceptance!- Lack of transparency leads to scepticism

● Documentation of model- Experiments should be described so others can repeat them- Same goes for models – but often not the case

From Schmolke et al 2010; TREE 25: 479-486

30

Outline

● Introduction● Uses of modelling within Syngenta● Uses of modelling in pesticide regulation● What enables the use of modelling in regulation?● Opportunities

31

Lord Kelvin

“I often say that when you can measure what you are speaking about and express it in numbers you know something about it; but when you cannot express it in numbers your knowledge is of a meagre and unsatisfactory kind: it may be the beginning of knowledge but you have scarcely, in your thoughts, advanced to the stage of science, whatever the matter may be.”

We need to quantitatively simulate processes resulting in

toxic outcomes - systems biology models, human disease

models, adverse outcome pathway (AOP) modelling

● From the Descriptive to the Predictive

32

chemical properties

in vitro data

in vivo data

Opportunities for prediction (QSARs and structural alerts)

33

Toxicologydatasets

Machinelearning

Questions

Models

34

Working towards a more collaborative approach

● Pre-competitive data sharing, the power of numbers, 3Rs● Influence of crowdsourcing, social media, open-source

Public domain data & models

Syngenta

data & models

Company Bdata and models

Company A data and models

35

Why are models used? Because they are useful!

So make sure your models are useful