What have we learnt about using HT screening as a source of agrochemical

leads?

John Delaney

Pharmaceuticals v Pesticides

Both interact with the same sorts of target – e.g. enzyme, receptor

Different economics

Pharmaceuticals – what price a life or a quality of life?

Pesticides – what price a bushel of wheat?

Quantities also differ somewhat…

A typical pharmaceutical delivery system

and another…

A typical pesticide delivery system

In vivo screening

In vivo high throughput screening

In vitro high throughput screening

                                 

                         

            

The short answer…

The quality of the chemical input to a screen matters

Proper analysis of the screen results matters

Logistics/cycle times matter

What do we mean by input quality?

For the purposes of this talk I won’t cover sample integrity, important though that is

The guiding principle is …

“If this compound were to hit on my screen, would I consider it a lead worth working on?”

If the answer is no, why did you screen it in the first place?

What do we look for in a lead (beyond potency) ?

A lead is a hit that is …

1. Novel

2. Distinct

3. Interesting

Novel

Is this compound similar to something I already know a lot about?

There are no prizes for re-discovering a well worked area of chemistry

We look for compounds that are dissimilar to anything in our corporate database – this assumes that we know everything there is to know about our corporate database!

Distinct

Are the compounds in the collection you’re testing different from each other?

A bunch of similar hits might only constitute one lead area

Every slot taken by an close analogue is a slot that could have been used to try a different area of chemistry

Interesting

Easy (and worthwhile) to define ‘uninteresting’

Non-specific, toxic crap – e.g. organo-mercurics, acid chlorides, nitro-phenols

Compounds with poor physical properties

Harder to define ‘interesting’ – what makes a compound ‘agchem-like’ ?

Interesting = Right physical properties?

Bioavailability – a combination of potency, stability and mobility

All three affected by the physical properties of the molecule

We know that certain combinations of phys props severely compromise mobility

We know that the presence of certain chemical groups can affect stability

Physical properties of agrochemicals

Not so very different from Lipinski’s ‘rule of five’

MWT between 200 and 500

clogP < 4

Basic pKa < 9 (big difference from pharma)

H-bond donors (OH,NH) < 3

Mobility – like Lipinski refers to passive transport only (Colin Tice, Pest Manag Sci 57:3-16 (2001)

How do we ensure that our input is good?

Apply rigorous filters to compounds we buy in

Designing decent properties directly into our own libraries

Encouraging signs that this is working – more leads from the same number of hits

We can cope with collections offered in a variety of formats – individual, plates or whole collections

Analysis

Analysis of hits traditionally done ‘by hand’

This becomes difficult as the number of screens and the number of compounds fed through them rise

Automation and standardisation part of the answer

Standardisation

Is each assay a unique case? Really?

Recording and storing data in a standard form greatly eases the task of developing analysis tools

Expect some up-front grief…

How do we analyse a bunch of hits?

Grouping similar structures together

Pulling relevant data from other sources

Turning raw biology into breakpoints

Flexible display of structures, activities and physical properties

Clustering

We tend to use Daylight substructural fingerprints as our molecular descriptor, the Tanimoto coefficient as our measure of similarity, and modified Jarvis-Patrick non-hierachical cluster analysis to group compounds – since you ask…

Unashamedly chemistry driven!

Groupings tend to chime with chemists’ intuition

Excel as a tool for data analysis

Ubiquitous – this is the way our chemists do most their data analysis

May not be the best tool for doing this kind of work, but…

Its short-comings can be addressed through programming effort

Take a general purpose tool and make it specific

The D1 batch system

Batch screening of compounds on in-vitro targets

A framework for collating data, analysis and driving analogue acquisition

Excel based – familiar to chemists

Incorporates clustering and data visualisation using AVS

Keeps track of what was done when and why

Cycle times

Cycle times can be surprisingly long

Often leads from the first stage of screening need ‘amplifying’

Rapid follow-up with analogues key

We would like library design to be an iterative process – delays in getting results compromise the effectiveness of this

Effects of changes to compound selection procedures

Analogue ordering

Potential for chaos here

Centralise the actual ordering process

‘Supersearch’ searches a hierarchy of databases and automatically eliminates duplicate compounds

Automatic annotation of database – “why was this compound ordered from Maybridge?”

Ordering done by adding an MFCD number to a spreadsheet

Conclusions

Good chemical input doesn’t just happen, you’ve got to work at it

Analysis can be made easier and faster – automate where possible, but consider the people doing the analysis

Cycle times are still a worry – some progress with making analogue ordering easier

And remember…

You’re only as good as your weakest link