Back Rapid lead compounds discovery through high-throughput screening
A_Pope_RQRM_LeadDisc_June_2016
1. Andy Pope
Discovery Partnerships with Academia (DPAc)
GlaxoSmithKline
RQRM 6ème colloque annuel
McGill University, Montréal, Québec
June 6th 2016
Lead discovery;
A critical step in the
development of innovative
new medicines
2. Topics
Current hit identification approaches (and philosophies)
A Short history of Diversity screening and current status
The importance of compound quality
A sampling of trends in screening
3. Target Identification
& Validation
Reagent &
Assay
Development
Hit Discovery Hit to Lead
Lead
Optimization
Activities;
• Identify potential
disease-linked target(s)
Methods;
Target validation;
- Genome sequence
data
- Tissue/cellexpression
- Literature search
- Expression modulation
Target tractability;
- Experience with similar
targets
- Target knowledge
e.g. modeling reveals
binding pockets, natural
modulators
Activities;
• Create materials
needed to support
hit discovery and
beyond
Methods;
- Expression cloning
- Protein tags
- BacMam cellular
expression
- Homogeneous assay
methods
Activities;
• Identify compounds
which modulate the
target in a desirable
way
Methods;
Screening;
- Knowledge-based
- Diversity (HTS, ELT)
- Focused sets
- Fragment-based
Chemical clustering
Screening informatics
Activities;
• Select and explore
promising chemical
series to find those
suitable for Lead
Optimization
Methods;
- Selectivity/specificity
assays
- Cellular assays
- Compound MOA
- SAR expansion
- Early safety assays
(e.g. hERG, p450, cell
health)
- Ligand efficiency
- IP potential
Activities;
• Optimize chemical
series to have
appropriate properties
to be a potential
medicine
Methods;
- SAR assays (selectivity,
orthology)
- Broad cpd profiling
- Cellular activity
- Pre-clinical models of
disease
- DMPK, regulatory
safety assays
- Cpd scale-up, cost of
goods
- IP secured
Commit to
approach
Commit to
target
Commit to
Lead series
Select clinical
candidate
From; The Role of Chemical Biology in Drug Discovery. Wiley Encyclopedia of Chemical Biology, Pope AJ (2012)
Hit Identification in the context of Drug Discovery
4. What constitutes a good target for a new medicine?
Presentation title 4
TARGET
VALIDATION
TARGET
TRACTABILITY
e.g.
Evidence for the role of target in
disease (e.g. genetics)
Evidence that pharmacological
manipulation will provide benefit
Understand potential safety issues
with approach
Current therapeutic approaches –
evidence that new approach will be
superior
e.g.
What is the best therapeutic
modality (i.e. small molecule vs
biopharm)?
For small molecules;
Evidence that the useful compounds
are likely to be found and can be
delivered to the site of action
- modality (e.g. inhibit vs activate)
- existing pharmacology
- target class
- potential drug binding sites in silico
5. Hit identification approaches (and philosophies)
Presentation title 5
REDUCTIONIST
HOLISTIC
COMPLEXITY
USE OF SPECIFIC KNOWLEDGE OF TARGET
TO DEFINE SCREENING SET
HIGHLOW
HTS/uHTS
Encoded
Libraries
Intact animal/
patient
Primary cell
Structure
based
design
Focused
screening
Re-
purpose
screening
Fragment
screens
phenotypic approaches
cell-based screens
biochemical screens
In-silico
design
Cell line
Membrane
Protein
Soluble
Protein
6. Hit identification approaches (and philosophies)
Presentation title 6
REDUCTIONIST
HOLISTIC
COMPLEXITY
USE OF SPECIFIC KNOWLEDGE OF TARGET
TO DEFINE SCREENING SET
HIGHLOW
HTS/uHTS
Encoded
Libraries
Intact animal/
patient
Primary cell
Cell line
Soluble
Protein
Structure
based
design
Focused
screening
Re-
purpose
screening
Fragment
screens
phenotypic approaches
cell-based screens
biochemical screens
In-silico
design
Time and labor intensive
Risk often in enabling systems
Success high if enabled
(i.e. ligand structures solved) Membrane
Protein
7. Hit identification approaches (and philosophies)
Presentation title 7
REDUCTIONIST
HOLISTIC
COMPLEXITY
USE OF SPECIFIC KNOWLEDGE OF TARGET
TO DEFINE SCREENING SET
HIGHLOW
HTS/uHTS
Encoded
Libraries
Intact animal/
patient
Primary cell
Structure
based
design
Focused
screening
Re-
purpose
screening
Fragment
screens
phenotypic approaches
cell-based screens
biochemical screens
In-silico
design
Opportunistic; based on known
properties of test compound set
Enriched for target class (e.g.
protein kinases) or compound type
(e.g. marketed drug sets)
Success moderate; knowledge is
rarely directly related to specific
new target
Cell line
Membrane
Protein
Soluble
Protein
8. Hit identification approaches (and philosophies)
Presentation title 8
REDUCTIONIST
HOLISTIC
COMPLEXITY
USE OF SPECIFIC KNOWLEDGE OF TARGET
TO DEFINE SCREENING SET
HIGHLOW
HTS/uHTS
Encoded
Libraries
Intact animal/
patient
Primary cell
Structure
based
design
Focused
screening
Re-
purpose
screening
Fragment
screens
phenotypic approaches
cell-based screens
biochemical screens
In-silico
design
Emphasize coverage of diversity
Recognize lack (limitations) of
knowledge of what binds
Success variable
(
Cell line
Membrane
Protein
Soluble
Protein
9. Diversity Screening Methods – High Throughput
Screening
Often the first line approach for GSK and other Pharma
Building & maintaining infrastructure represents a very large investment -
Build the best possible library and screen as many compounds as possible
HTS within Pharma companies – typically 1-2M compounds
Academic HTS – typically 10-300K compounds
The more novel the target, the less is known >> HTS preferred option
Presentation title 9
10. A Short History of HTS
Presentation title 10
1990’s HTS 96/384-well, mixtures, slow/unreliable automation
2000’s ultra-HTS 1535-well, singles, collection growth, focus on assay
and process quality/speed
2010’s Increase disease relevance, focus on hit quality, new
diversity methods (e.g. ELT)
11. Typical Modern HTS process
Presentation title 11
HTS Assay Protocol and Scaled Reagents
Validation & Pre-production
Screen Peer Review
Primary Screen
Confirmation of Actives
Cheminformatic Analysis
Dose – response Testing
Screen Output Review
Robust assay able to detect desired pharmacology
- 10K validation X3
- 100K pre-production
-Ensure assay quality
- Plan for screen and triage
- 2M cpds @ 10 mM
- Typically 5-10 days
- 10 mM X2
- Up to 20K cpds
- Chemical clustering
- Sample active diversity
- 11 point DR curve X2
- Up to 4K cpds
- Summarize output
- Plan for hit triage
Hit Triage/qualification
12. 15+ Years of HTS screening
Cellular/Biochem ModeTarget class
>350 HTS campaigns of >1M cpds within GSK alone
Success rates vary considerably; some clear trends
Heuristics from huge data volume (~109 data points)
e.g.
- Assay technologies
- Role of chemical properties
- Nuisance effects
- Predict success on a protein/interaction class basis
13. Building an HTS compound set
13
Hit
Non-Hit
Lead
multiple exemplars per “cluster”
7%
30%40%
23%
Lipophilicity,cLogP
Molecular Weight (Da)
As large as economics can support
High quality (i.e. LCMS validated) cpds
Good chemical property space occupancy
Supply chain automation/technology is critical
Majority of compounds sourced externally
14. Compound quality is critical for success
Poor compound physicochemical properties are quite strong predictors of failure in
drug discovery
Lipophilicity Promiscuity (safety)
Low solubility Multiple issues (e.g. bioavailability, formulation)
Unfavorable properties are often difficult to engineer out with chemistry
Industry-wide move to improve the quality of starting points >> clinical candidates
15. ClogP
%CpdsinClogPBin
Cumulative%Cpds
Middle 80% of Cpds
1 5
ClogP
HitRate(%)
1.14%
3.31%
4.5%
1.1%
Overall hit rate rises ~3-fold across the middle 80% of the screening deck
Biases screening towards selection of poorest quality compounds
Large variations in effect from screen to screen
- bins containing 1M or more records across 350+ HTS are shown
HTS can be biased towards poor quality hits - lipophilicity
17. HTS Hit Promiscuity and Chemical Properties
*Effect frequency Index (EFIX) = % of screens where cpd yielded >X.RSD effect, where total screens run =>100
Compounds hitting
~1-2 targets
“Dark matter”
Compounds hitting
>10% of targets
PropertyForecastIndex*
%EFI3* 0 1 2 3-5 5-10 10-20 >20
*Property Forecast Index = ChromLogD7.4 + #AromRings
18. Hit Quality, Hit Qualification & Target Tractability
HR.Yobs = HR.YtargetReal + HR.YtargetArtefact + ∑ HR.Y*system1…n
HR.Yobs = Observed % of library Y samples which yield activity
HR.YtargetReal = % of library Y which yields a productive pharmacological
effect
HR.YtargetArtefact = % of library Y for which tested cpd samples create a
(reproducible) artefact
∑ HR.Y*system1…n = Sum of all sources of system errors resulting in false
positive signals in a screening assay
HR.Yobs
HR.YtargetRealLow
High
High
Low
Increasing false
positives
Increasing false
negatives
High
tractability
Low
tractability
Low
tractability
?
Optimize assays to maximize proportion of real hits
Recognize that the “real” hit rate is often extremely low
- so most hits may be artifacts
Hit qualification/disqualification methods are crucial
19. Hit Quality, Hit Qualification & Target Tractability
HR.Yobs = HR.YtargetReal + HR.YtargetArtefact + ∑ HR.Y*system1…n
HR.Yobs = Observed % of library Y samples which yield activity
HR.YtargetReal = % of library Y which yields a productive pharmacological
effect
HR.YtargetArtefact = % of library Y for which tested cpd samples create a
(reproducible) artefact
∑ HR.Y*system1…n = Sum of all sources of system errors resulting in false
positive signals in a screening assay HR.Yobs
HR.YtargetRealLow
High
High
Low
Increasing false
positives
Increasing false
negatives
High
tractability
Low
tractability
Low
tractability
?
Optimize assays to maximize proportion of real hits
Recognize that the “real” hit rate is often extremely low
- so most hits may be artifacts
Hit qualification/disqualification methods are crucial
Minimized by high levels of
quality and process control
Minimized by annotation of
cpd samples which cause
nuisance effects and
improvement in assays
20. Hit Qualification vs. Disqualification
Elapsed time
Elapsed time
Elapsed time
ScreenActives(~103)
Primaryscreen(~2x106)
ScreenActives(~103)
Primaryscreen(~2x106)
ScreenActives(~103)
Primaryscreen(~2x106)
A. Typical process – Hit Disqualification paradigm
B. Hit Qualification paradigm – Target with high chemical tractability
C. Hit Qualification paradigm – Target with low chemical tractability
Disqualificationassays
-e.g.interference,
redox,orthogonal
DPU chemists
Select templates
of interest100’s
1000’s
1- 10’s
Pre-Chemistry
•Available Analogs
• PropertySAR
Biology
• Selectivity
• Biophysics/MOA
• Cellular assays
DPU H2L Chemistry
• Explore SAR
•Finalize template(s)
Biology
• Disease relevance
Leadseriesor OR ……again
Qualifyhits
(e.g.ASMS)
CIX
Biology
• Selectivity
• Biophysics/MOA
• Cellular assays
H2LChemistry
• Explore SAR
•Finalize template(s)
Biology
• Disease relevance
MulitipleLead series
Pre-Chemistry
• Re-purification
• Re-synthesis
• Available Analogs
Qualifyhits
(e.g.ASMS)
Lead?
Targettractable? OR….
Use other approaches?
100’s
1000’s
1000’s
1-10’s
10’s
21. Hit qualification - rapid focus on “real” hits
- Can also reveal target chemical tractability
e.g. Protein Kinase X; ~3% HR 92% of hits qualify by ASMS
e.g. Epigenetics Target-Y; ~0.5% HR, 3% of hits qualify by ASMS
ClogP
MW
• = positively binder
• = active in HTS
assay, no binding
Poor chemical tractability
Excellent chemical tractability
MW
ClogP
Allen Annis et al. Current Opinion in Chemical Biology 2007, 11:518–526
22. HTS has limitations as a diversity screening method
- addressed by Encoded Library Technology (ELT)
Infrastructure and screening costs limit collection size - quantal scaling not possible
- lead-like chemical space is huge (>1020)
HTS campaigns can only use one set of highly defined reagents and conditions
- limits ability to screen under a range of conditions, one target per screen
High level of commitment to target needed to justify investment in target
Encoded Library
Technologies (ELT)
e.g. Compound storage infrastructure;
HTS – 2M compounds
ELT – 1B compounds
e.g. Protein requirements;
HTS – > 10 mg’s of protein (3 million wells)
ELT - < 100 mg’s of protein (1 set of selections)
23. ‘Split and pool’
synthesis
ligate
(96 tags)
chemistry
(96 different BB’s)
Precipitation
(deprotect & purify if
necessary)
96-well plate
Chemistry
Encoding
Encoding
Cycle 1: 96 unique compounds
Cycle 2: 96 x 96 (9,216) unique compounds
Cycle 3: 96 x 96 x 96 (884,736) unique
compounds, each with a DNA ‘barcode’
SPLIT
ELT - Library Synthesis Strategy
POOL
24. ELT Selection and Hit Confirmation Process
24
Step 4:
Sequencing by Illumina HiSeq
(1-2 weeks)
Step 2:
Affinity Selections
(1 week)
Step 1:
Target immobilization and
activity confirmation
(1 week)
Affinity
Matrix
Step 3:
Sample preparation/
amplification for sequencing
(2 days)
TGTCTCCACCC AGTGGTGTGAG GTCACGGTCCA GACCTCCATTC TT
ACAGAGGTGGG TCACCACACTC CAGTGCCAGGT CTGGAGGTAAG AA
N
N
N
N
H
N
N
O
O
O
3
4
2
Each BB is encoded by
a unique tag set. In
most cases multiple
tags are used for each
building block
NH2
O
N
O
O
N
CAAGTCCTTGTACCACGAAGAGCTGGT
ACGTTCAGGAACATGGTGCTTCTCGAC
CAAGTCCTTGTACCACGAAGAGCTGGT
ACGTTCAGGAACATGGTGCTTCTCGAC 2
4
3
1
Cycle 1 Cycle 2 Cycle 4
NH
N
O
OH
N
H
1
2
4N
N
N
N
H
N
N
R
NH2
O
O
O
CH3
CH3
N
O
NH2
O
O
O
CH3
CH3
OH
Each building block is
encoded by a unique and
degenerate set of “codons”
Step 5:
Translation of sequences
into structures
(1-2 weeks)
Step 6:
Off-DNA synthesis and hit
evaluation
(2-3 weeks per chemotype)
26. ELT and HTS by the numbers
Presentation title 26
DNA-Encoded Library (DEL) Catalog
MDR-Waltham
Updated May 12, 2015
ELT HTS
130 libraries to date >2M HTS deck
> 1B warheads ~400K chemotypes
ELT HTS
~ 600 targets ~500 targets
70M binding events ~500K initial actives
ELT HTS
300M sequences/wk 2M assay wells/wk
20M sequences/target 3M assay wells/target
27. Increasing the relevance of screening assays
Phenotypic Screening
Presentation title 27
http://www.sbpdiscovery.org/technology/sr/Pages/LaJolla_HighContentScreening.aspx
Screen for new hits via impact on cellular
processes, without knowledge of specific
target
- Full HTS possible (but challenging)
- Targets presented in correct context (e.g.
complexes
- Screen selects for cell-penetrant cpds
- De-convolution of mechanism needed?
- Chemistry challenging
Disease-relevant assays to validate and
optimize leads from other methods
- Primary (+iPS derived) cells
- Little/no manipulation (e.g. over-expression)
- Maximize translation to in vivo/patient
28. Can target tractability be assessed earlier?
Presentation title 28
TARGET
VALIDATION
TARGET
TRACTABILITY
Current Paradigm;
Select and validate a target, only REALLY know if it is tractable after investing in screening
TARGET
TRACTABILITY
Alternate approach;
Select a number of potential targets (e.g. pathway members, protein class, essential bacterial)
Screen in parallel using low-cost approach
Select targets based on results – starting with good tractability
TARGET
INVESTMENT
CANDIDATE
TARGETS
TARGET
TRACTABILITY
TARGET
TRACTABILITY
TARGET
TRACTABILITY
TARGET
INVESTMENT
29. ELT Panel Screening and for early tractability assessment and
probe discovery
Manuscript in preparation
Panel of target proteins
Binding
Confirmation
Hits/leads
Tools/probes
Therapeutic
Opportunities
Priority for
synthesis
Targets
with signal
Targets
with MoA
Amenable
to selection
All targets
Tractable targets
29
Express/purify
ELT selections
Feature analysis
30. ELT Panel Example – Revisiting essential bacterial targets
Presentation title 30
Drugs for bad bugs: confronting the challenges
of antibacterial discovery. Payne, D.J. et al. (2007),
Nature Drug Discovery 6: 29-40.
82 > Expression
73 > Tagged
Proteins
70 > ELT
selections
57 > Sequencing
/ Analysis
9 targets removed
- poor expression
3 targets removed
- poor purification
25 >
Chemistry
17 > MIC
4 > MoA
Jan ‘13 DecOctFeb Mar Apr May Jun Jul Aug Sep Nov Jan Feb 14
Analysis & PSC
Planning
Chemistry MIC
panel initiated
MoA Studies
Initiated
MST studies
for on target
hits
Selections &
Sequencing
Construct Synthesis
82
Targets
Chosen
Genome
Sequence
Analysis
30
13 targets removed
- poor capture
Tools/leads with validated anti-bacterial mechanism
Historical experience from 70
HTS campaigns;
Essential genes in S. aureus
32 targets removed
- lack of feature
enrichment
31. Conclusions
Disease-validation and Chemical tractability are equally important for successful (small
molecule) Drug Discovery
For most new targets, diversity screening methods are the best option
Tractability is often difficult to predict up-front (anecdotal drug-hunter experience factor!)
HTS and ELT provide complementary methods, but ELT is more scalable. In both cases,
major investment is required in order to ensure overall success
Quality must be a major focus in screening. Process quality is largely solved, hit quality
(and qualification) is a work in progress
A drive towards more disease relevant assays is underway and more can be expected in
future (e.g. CRISPR engineered systems, primary cells, organoids etc.)