CINF 29: Visualization and manipulation of Matched Molecular Series for decision support
1 like3,633 views
The document discusses the visualization and manipulation of matched molecular series to aid decision support in drug discovery. It explores the concept of matched pairs and series, highlights their application in predicting physicochemical property changes, and discusses a dataset-centric approach for generating predictions based on molecular activities. The proposed visual interface aims to encourage exploration and hypothesis generation while addressing the limitations of traditional matched pair methodologies.
![Matched (Molecular) Pairs
[Cl, F]1.6
3.5
Coined by Kenny and Sadowski in 2005*
Easier to predict differences in the values of a property
than it is to predict the value itself
* Chemoinformatics in drug discovery, Wiley, 271–285.](https://image.slidesharecdn.com/visualizationmatsy-150825090607-lva1-app6892/85/CINF-29-Visualization-and-manipulation-of-Matched-Molecular-Series-for-decision-support-2-320.jpg)
![Matched Series of length 2
= Matched Pair
[Cl, F]
“Matching molecular series” introduced by Wawer and
Bajorath, J. Med. Chem. 2011, 54, 2944](https://image.slidesharecdn.com/visualizationmatsy-150825090607-lva1-app6892/85/CINF-29-Visualization-and-manipulation-of-Matched-Molecular-Series-for-decision-support-4-320.jpg)
![Matched Series of length 3
[Cl, F, NH2]](https://image.slidesharecdn.com/visualizationmatsy-150825090607-lva1-app6892/85/CINF-29-Visualization-and-manipulation-of-Matched-Molecular-Series-for-decision-support-5-320.jpg)
![Ordered Matched Series of length 3
3.5
2.1
1.6
pIC50
[Cl > F > NH2]](https://image.slidesharecdn.com/visualizationmatsy-150825090607-lva1-app6892/85/CINF-29-Visualization-and-manipulation-of-Matched-Molecular-Series-for-decision-support-6-320.jpg)
CINF 29: Visualization and manipulation of Matched Molecular Series for decision support
- 1. Visualization and manipulation of Matched Molecular Series for decision support Noel O’Boyle and Roger Sayle NextMove Software 250th ACS National Meeting, Boston 16th Aug 2015
- 2. Matched (Molecular) Pairs [Cl, F]1.6 3.5 Coined by Kenny and Sadowski in 2005* Easier to predict differences in the values of a property than it is to predict the value itself * Chemoinformatics in drug discovery, Wiley, 271–285.
- 3. Matched Pair usage • Successfully used for: – Rationalising and predicting physicochemical property changes – Finding bioisosteres • Not very successful in improving activity – Activity changes dependent on binding environment • Need to look beyond matched pairs
- 4. Matched Series of length 2 = Matched Pair [Cl, F] “Matching molecular series” introduced by Wawer and Bajorath, J. Med. Chem. 2011, 54, 2944
- 5. Matched Series of length 3 [Cl, F, NH2]
- 6. Ordered Matched Series of length 3 3.5 2.1 1.6 pIC50 [Cl > F > NH2]
- 7. The Matched Pair mentality • There can only be two – Like inhabitants of Flatland ignorant of a third dimension • What is the equivalent of pair for three? – Triad, trio, triple? • A matched pair represents a transformation from A->B – How would that work if it there were three?
- 9. Matched Series have prefeRred orders Series Enrichment Observations Br > Cl > F > H 5.36* 256 Cl > Br > F > H 3.14* 150 H > F > Cl > Br 1.53* 73 Br > Cl > H > F 1.40 67 F > Cl > Br > H 1.36 65 Cl > F > Br > H 0.96 46 … … … H > F > Br > Cl 0.77 37 … … … H > Br > F > Cl 0.48* 23 Cl > H > F > Br 0.48* 23 Cl > F > H > Br 0.48* 23 H > Cl > F > Br 0.42* 20 Br > F > H > Cl 0.40* 19 F > H > Br > Cl 0.40* 19 H > Cl > Br > F 0.38* 18 F > Br > H > Cl 0.36* 17 Br > H > F > Cl 0.17* 8 The fact that certain orders are preferred may be used as the basis of a predictive method
- 10. Find R Groups that increase activity A > B Query matched series A > B > C C > A > B D > A > B > C D > A > C > B E > D > A > B … R Group Observations Obs that increase activity % that increase activity D 3 3 100 E 1 1 100 C 4 1 25 … … … In-house O’Boyle, Boström, Sayle, Gill. J. Med. Chem. 2014, 57, 2704.
- 11. The Dataset-Centric approach • “Here is my dataset of molecules with activities – now tell me what to make next” • Pro: – Easy for users to get up-and-running – Fits with their existing way of thinking • Don’t need to think too much about matched series • Con: – User is one step removed from the matched series data on which the predictions are actually based – Dataset is fixed: cannot play with around with the prediction input
- 12. Goals for the interface • Visual interface based around R-Groups as first- class objects arranged in ordered series – Promote new paradigm – Make it clear that the scaffold is not involved • Should help break the “matched pair” mentality – Just a particular case of matched series • Should be easy to play with – Easy to manipulate and quick to respond
- 13. • Drag-and-drop R Groups into slots to represent observed activity order – The query matched series
- 15. Pros • Easy to play around with – Swap around order of R groups – See what happens if you follow the predictions • May suggest hypotheses • Useful for searching (not just for predictions) • Tablet-friendly ConS • The user needs to be able to provide an ordered matched series as a query – You can’t just provide a dataset of molecules
- 16. No Chemistry required • Predictions are solely based on the order of R groups in a matched series – Not using any calculated properties • Images of all R groups in ChEMBL can be generated in advance (~65K) • ⇒ A cheminformatics toolkit is not required for the interface or even for making predictions • In practice, we do use a toolkit to allow the user to enter R groups as SMILES
- 17. Use Case #1 Are matched Series Predictions symmetric?
- 18. Are matched series predictions symmetric? • If A>B>C>D is a highly preferred order – Then D>C>B>A also tends to be preferred • Hypothesis: – if A reduces the activity given D>C>B ⇒ it will also improve the activity given B>C>D • If true, then we have twice as much data to use for predictions – Let’s find out….
- 19. ? > Cl > F > H H > F > Cl > ?
- 20. Use case #2 Topliss Decision Tree
- 21. Topliss Decision Tree Topliss, J. G. Utilization of Operational Schemes for Analog Synthesis in Drug Design. J. Med. Chem. 1972, 15, 1006–1011.
- 24. ChEMBL-BASED Decision Tree (One of Many)
- 25. http://nextmovesoftware.com noel@nextmovesoftware.com @nmsoftware Visualization and manipulation of Matched Molecular Series for decision support