Development of web-based Chemistry Design Tools at GlaxoSmithKline
Stephen PickettMay 20th, 2015ChemAxon UGM, Budapest
The Discovery Process
gene protein
targetscreen and
identify lead
Lead
optimisation
chemical
diversity
(compound
library)
test safety& efficacy
in animals and
humans
Targets Hits Leads Candidates Drugs Products
Drug discovery and computational design and prediction
Target
Selection
Candidate
Selection
Product
Differentiation
Market
Entry
Target IDLead
DiscoveryLead
Optimisation
Preclinical
Development POCFull
DevelopmentRegistration
Predictive ADMET
Structural Genomics
Sequence homology
Molecular Diversity
Library Design
Molecular Similarity
HTS analysis
(Q)SAR analysis
Attrition analysis/risk assessment
Structure Based Design
Ligand based Design
Solid state/multi scale modelling
Drug discovery made simple
Corporate
collection
Screen
Analysis
What to
make next?
Make it
What’s here to begin
with? What should we
add?
Primary target assay
Selectivity assays
Developability screens
Disease models
“Decision support”
Clustering
Machine learning
Visualisation
Are we there yet?
Predictive models
Design algorithms
Current State: tools covering many application areas
• Data access and reporting
• SAR analysis and visualisation
• Reagent selection and enumeration
• Similarity, clustering, focussed sets
• Alignment and pharmacophore generation
• Protein/ligand visualisation and interpretation
• QSAR modelling
– Local and global models
• Docking, scoring, idea prioritisation
55/22/2015
New compound design
Email alerting service
Configurable data refresh intervals
Launch of data into excel
or Spotfire templates
Compound property tracking
Project Data Management - CornerShops
Project Data Management – Structure Viewers
Crystallography data and
comments
Cornershop data
Prealigned view
GRID and other map
views
Electron
density
BioDig ADME (Hussain, Rea, JCIM, 2010, 50, 339-348)
“What possible changes could be made to the
compound to improve its clearance ?
Binned ACTIVITY_CHANGE
1
2
3
5
4
17
21
7
2
-2.2 -1.8 -1.4 -1 -0.6 -0.2 0.2 0.6 1 1.4 1.8 2.20
5
10
15
20
PROFILE_CLUSTER - 22
PROFILE_CLUSTER - 28
PROFILE_CLUSTER - 33
PROFILE_CLUSTER - 23
PROFILE_CLUSTER - 31
PROFILE_CLUSTER - 34
PROFILE_CLUSTER - 27
PROFILE_CLUSTER - 32
PROFILE_CLUSTER - 38
- - - 0 + + + - - - 0 + + + - - - 0 + + +
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
70
80
90
100
Matched Molecular Pairs analysis in compound design
State of the art design in global tools
MW HBD RB HBA CELL SIZE
0
100
200
300
400
0 192 384 576 768 960 1152 1344 1536
Library Size
Occu
pie
d C
ells
CONSTRAINED
UNCONSTRAINED
SELECT
Div
ersi
ty
Library Size
T. Wright, V. J. Gillet, D.V.S. Green & S.D. Pickett, Optimizing the Size and Configuration of
Combinatorial Libraries. J. Chem. Inf. Comput. Sci. (2003), 43(2), 381-390.
QSAR modelling and deployment – expert to end user (JCAMD, 2013, 27, 321-336)
Published Model available as WS
To this…
GREAT look and feel
No compromiseScience
Evolving scienceCSC plug-ins
Replace & retireAdept, Adamantis, CSC web apps
Robust
High PerformancePersonalisediDesign
ConfigurableiDesign Fragments
CollaborativeIdeas sharing
Target IC50 IC50 AVG IC50
Target1 1.99 -1.84 0.08
Target2 2.50 -0.10 1.20
Target3 1.81 4.62 3.22
Target4 5.91 2.59 4.25
Target5 4.15 4.52 4.34
Target6 4.22 4.58 4.40
Target7 4.34 4.49 4.41
Target8 5.11 4.50 4.80
Target9 6.31 5.40 5.86
Target10 5.37 6.47 5.92
Target11 6.48 7.22 6.85
Target12 8.43 9.19 8.81
Target13 9.31 11.07 10.19
Target14 11.89 11.72 11.81
Target15 10.88 14.74 12.81
Target16 11.53 14.31 12.92
Target17 12.62 13.25 12.93
Target18 14.58 13.36 13.97
Target19 13.34 15.34 14.34
Target20 14.13 15.23 14.68
MW 451 fSP3 0.21
cLogP 4.1 TPSA 77
HBD 1 HBA 3
SKETCH
ENUMERATE
DOCK
PLOT
IMPORT
Integrated, Intuitive2D, 3D, data, visuals
The concept originated from our desire to improve efficiency,
increase compound quality, and reduce attrition
Efficiency
• Drug design
• Selection of chemistry targets
Quality
• Physico-chemical properties
• PK/Safety
Success
• Reduce attrition
• Patient benefit
How to achieve our goals: what do our chemists want and need?
• 12 Workshops
– 4 R&D sites
– ~100 chemists
• Open session at the GSK Chemistry Conference (including our CEO!)
• Pan R&D Survey
– Who designs molecules?
– How?
5/22/2015 15
What is LiveDesign?
• A design tool for small molecule medicinal & computational chemists– Drawing, importing, enumerating structures
– Calculating properties
– Searching structure databases
– Import of measured data
– Visualisation of data & structures
– Generation & visualisation of 3D conformers
– Visualisation of Protein structures
– Collaboration environment
– Project based views & preferences
– Ability to add new calculations & workflows
– Advanced functionality (…see later)
• Consolidation of many existing tools into a single platform, replacing those that
are outdated or unsustainable
225/22/2015
Predefined templates
Sketch, search, import, enumerate
Collaboration
Design and analysis
Models and computational tools
Protein data and experimental results
Advanced tools
More features being developed and integrated……
• Many more…….Prioritization needed by the chemistry community over the next 1-2 months
5/22/2015 27
Binding Interaction HighlightsP450 Site of Metabolism
Multi-Parameter Optimization
Model Versioning
Atomic Contribution Mapping
Benefits
5/22/2015 28
• Reduce cycle times for lead optimization
• Unlock users
• EfficiencyUnified platform
• Get peer review and team feedback on both virtual and synthesized compounds
• Share ideas and contribute to projects across GSK
Collaborative environment
• Time and cost savings
• The cost of losing a compound after Candidate Selection estimated as £6M
Simplification
Challenges!!!
5/22/2015 29
• Niche workflows
• Customized tools established
Heterogeneous end user groups
in a global organization
• Worldwide access
• Security
• BugsPerformance
• Co-existing software
• Browsers and plug-insCompatibility
Key factors to success
5/22/2015 30
• Speed similar or better than existing tools
• Smooth - not “buggy”Performance
• Appealing interface
• Quality renderingLook and Feel
• Easy to use
• Minimal trainingIntuitive
• GSK-Schrödinger
• End users
• Shape the final release
Communication
Get it right
Happy chemists!
High user uptake