Page 1
© 2017 Optibrium Ltd.Optibrium™, StarDrop™, Auto-Modeller™, Card View™ and Glowing Molecule™ are trademarks of Optibrium Ltd.
CINF Drug Discovery Cheminformatics Approaches
August 23rd 2017
Integrated Cheminformatics to Guide Drug DiscoveryMatthew Segall, Ed Champness, Peter Hunt, Tamsin Mansley
1
Page 2
© 2017 Optibrium Ltd.
Overview
• The impact of fragmentation of cheminformatics tools
• Challenges of integration
− Examples: Data access and docking/alignment
• Illustrative application
• Conclusions
2
Page 4
© 2017 Optibrium Ltd.
Fragmentation of Cheminformatics Tools
• Many capabilities are required to drive drug discovery projects− Database access
− Visualisation and analysis of data
− 2D predictive modelling (QSAR)
− 3D structure-based design
− …
• Disparate tools for these features create bottlenecks and inefficiencies− Reformatting of data to move between applications
o Time consuming and lost information
− Different user interfaces
o Training burden
− ‘Expert’ tools can be impenetrable to non-computational chemists
o Support from experts even for mundane tasks
o Delay and distraction of experts from adding scientific value
4
Page 5
© 2017 Optibrium Ltd.
Common Situation
6
MedicinalChemist
Database(s)(Internal and external)
ADMETpredictions
ComputationalChemist
Ligand-baseddesign
Structure-baseddesign
Visualisation
Excel Informatician
Pipelines
Page 6
© 2017 Optibrium Ltd.
Better Situation
7
MedicinalChemist Computational
ChemistInformatician Data server
Docking server
Model server
Page 7
© 2017 Optibrium Ltd.
Advantages
• Project leader can have all information available in one place
− (Ideally) single, common interface
− Can get instant feedback on new design ideas
• Computational and data experts do not waste time on routine, mundane calculations
− Focus on scientifically interesting and valuable activities
• Encourages closer collaboration
− Uninteresting ideas are quickly eliminated
− Collaborations can focus on high-value ideas for more detailed investigation
9
Page 9
© 2017 Optibrium Ltd.
The Challenges of Integration
• Different software preferences for many tasks
− Database/ELN providers
− ‘Favourite’ docking or alignment platforms
− Must be agnostic to the source of data and predictions
• Different architectures
− Web, web services, desktop, command line…
• Interaction is key
− Stimulates new ideas and strategies
− Black boxes are not trusted
• Must be user-friendly and intuitive
− Low barrier to use
− ‘Gluing’ lots of different software together leads to a poor user experience
11
Page 10
© 2017 Optibrium Ltd.
Example: Database accessChallenge
• Provide access to multiple data sources, e.g.− SQL via ODBC
− Flat files
− Web services…
• User-friendly definition of search criteria and fields− Support for criteria based on chemical structure, numerical, date, textual and
categorical fields
• Save, share, edit and execute pre-defined queries− Stored individually or by project
• Support for multiple data aggregation levels− Drill-down to underlying data
12
Page 11
© 2017 Optibrium Ltd.
Example: Database accessSolution
13
Wrapper
Flat filese.g. SD, CSV
Simple databasese.g. Oracle, MySQL
ODBC
Custom dataarchitectures
http://www.optibrium.com/community/videos/introduction-to-stardrop-modules-and-features/356-queryinterface
Drag and drop to add query criteria
Edit query terms in place
Load and save queries
Select data to retrieve
Combine criteria with ‘or’ and ‘and’
Select data source
Page 12
© 2017 Optibrium Ltd.
Example: Docking and AlignmentChallenge
• Enable non-expert users to run docking, pharmacophore or conformation
generation calculations
− Quick feedback on new compound designs
− Link information from 3D models with other analyses and data
• Support for all major 3D modelling platforms
− Flexibility to run models using preferred methods
• Expert computational chemists can easily publish new 3D models
− Configuration files for new targets ‘drop in’ on server
• Server to enable queuing and batch processing of long calculations
14
Page 13
© 2017 Optibrium Ltd.
Example: Docking and AlignmentSolution
15
GOLD…
Pose Generation
Server
Wrapper
WorkerWorker Worker… …
Pose Generation
Server
http://www.optibrium.com/community/videos/introduction-to-stardrop-modules-and-features/375-pgi
Page 14
Example Application
Page 15
© 2017 Optibrium Ltd.
Background
• Target: Heat-shock protein 90 (HSP90)
• Initial hits based on fragment based drug discovery
− PDB: 2XAB
− SAR explored around resorcinol in isoindoline resorcinol series
• Objectives
− Explore substitutions and replacements for isoindoline
− Identify high quality compounds for oncology target
− Decide on synthetic strategy
• Based on initial SAR from Woodhead et al. J. Med. Chem. 53 p. 5956 (2010)
− N.B. This is not the process used in this project; structures and data used for illustrative purposes only
17
Page 16
© 2017 Optibrium Ltd.
Retrieve Data for Latest HitsUnsubstituted isoindolines
18
Page 17
© 2017 Optibrium Ltd.
Initial SAR
19
Page 18
© 2017 Optibrium Ltd.
Initial SAR
20M.D. Segall et al. (2015) Drug Discov. Today 20(9) pp. 1093-1103
Page 19
© 2017 Optibrium Ltd.
Crystal StructurePDB 2XAB
21M.D. Segall et al. (2015) Drug Discov. Today 20(9) pp. 1093-1103
Page 20
© 2017 Optibrium Ltd.
Dock with HSP90 Model
22
Page 21
© 2017 Optibrium Ltd.
HSP90 Docking Results
23
Page 22
© 2017 Optibrium Ltd.
Search eMoleculesSubstituted isoindolines and potential replacements
24http://www.optibrium.com/community/videos/introduction-to-stardrop-modules-and-features/357-stardropemolecules
Page 23
© 2017 Optibrium Ltd.
Search eMoleculesResults
25http://www.optibrium.com/community/videos/introduction-to-stardrop-modules-and-features/357-stardropemolecules
Page 24
© 2017 Optibrium Ltd.
Clip Reagents
26M.D. Segall et al. (2015) Drug Discov. Today 20(9) pp. 1093-1103
Page 25
© 2017 Optibrium Ltd.
Enumerate Library
27
Page 26
© 2017 Optibrium Ltd.
Enumerated Library
28
Page 27
© 2017 Optibrium Ltd.
Dock Library
29
Page 28
© 2017 Optibrium Ltd.
Dock Library
30
Page 29
© 2017 Optibrium Ltd.
Good Docking Score is Not EnoughMulti-parameter optimisation
• Also require good Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) properties
31M.D. Segall (2012) Curr. Pharm. Des. 18(9) pp. 1292-1310
Page 30
© 2017 Optibrium Ltd.
MPO Scores
32M.D. Segall (2012) Curr. Pharm. Des. 18(9) pp. 1292-1310
Page 31
© 2017 Optibrium Ltd.
Compound Overview
33M.D. Segall (2012) Curr. Pharm. Des. 18(9) pp. 1292-1310
Page 32
© 2017 Optibrium Ltd.
Identify Chemotypes
34M.D. Segall (2012) Curr. Pharm. Des. 18(9) pp. 1292-1310
Page 33
© 2017 Optibrium Ltd.
Stack Chemotypes
35M.D. Segall et al. (2015) Drug Discov. Today 20(9) pp. 1093-1103
Page 34
© 2017 Optibrium Ltd. 36M.D. Segall (2012) Curr. Pharm. Des. 18(9) pp. 1292-1310
Page 35
© 2017 Optibrium Ltd.
Score Distribution
37M.D. Segall (2012) Curr. Pharm. Des. 18(9) pp. 1292-1310
Page 36
© 2017 Optibrium Ltd.
Compound Selection
38M.D. Segall (2012) Curr. Pharm. Des. 18(9) pp. 1292-1310
Page 37
© 2017 Optibrium Ltd.
Conclusion
• Integration of cheminformatics and computational chemistry tools is essential for efficient drug discovery
− Quick and good decisions on compound selection and design
• Ease-of-use and interactivity critical
− All users should be able to intuitively access data and predictions
− Bring together all data to target high-quality compounds
• Big challenges
− Compatibility with wide range of vendors and in-house platforms
− Support for variety of architectures
• For more information: www.optibrium.com
39
Page 38
© 2017 Optibrium Ltd.
Acknowledgements
• Optibrium team
− Too many to list!
• BioSolveIT – Access to FlexX and HSP90 model
− Christian Lemmen
− Marcus Gastreicht
− Carsten Detering
• Collaborators in development of query and pose generation interfaces
− Zoetis, The Edge, ChemAxon, IDBS…
− BioSolveIT, CCDC, CCG, OpenEye…
40