A Network Visualization of Structure Activity Landscapes<br />Rajarshi Guha<br />NIH Chemical Genomics Center<br />March 2...
Structure Activity Relationships<br />Similar molecules will have similar activities<br />Small changes in structure will ...
Exceptions Are Easy to Find<br />Ki = 39.0 nM<br />Ki = 1.8 nM<br />Ki = 10.0 nM<br />Ki = 1.0 nM<br />Tran, J.A. et al., ...
Structure Activity Landscapes<br />Rugged gorges or rolling hills?<br />Small structural changes associated with large act...
Structure Activity Landscapes<br />
Characterizing the Landscape<br />A cliff can be numerically characterized<br />Structure Activity Landscape Index (SALI)<...
Fingerprints<br />Lots of types of fingerprints<br />Indicates the presence or absence of a structural feature <br />Length...
Visualizing the SALI Matrix<br />
Visualizing SALI Values<br />Alternatives?<br />A heatmap is an easy to understand visualization<br />Coupled with brushin...
Visualizing the SALI Graph<br />Nodes are ordered such that the tail node in an edge has lower activity than the head node...
Varying the Cutoff<br />The cutoff controls the complexity of the graph <br />Higher cut offs will highlight the most signifi...
Varying Fingerprint Methods<br />Shorter fingerprints will lead to more “similar” pairs<br />Requires a higher cutoff to foc...
Varying the Similarity Metric<br />
Different Molecular Representations<br />The nature of the representation can significantlyaffect the landscape<br />SALI ma...
Different Activity Representations<br />Using the Hill parameters from a dose-response curve represents richer data than a...
SALI Curves from DRCs<br />No difference in major cliffs<br />Some of the minor cliffs are highlighted using the DRC inste...
Better Visualization - SALIViewer<br />http://sali.rguha.net<br />
Glucocorticoid Inhibitors<br />62 dihydroquinolinederivatives<br />IC50’s reported, some values were censored<br />50% SAL...
Glucocorticoid Inhibitors<br />62 dihydroquinolinederivatives<br />IC50’s reported, some values were censored<br />50% SAL...
Glucocorticoid Inhibitors<br />Moving from ally or phenylethyl to ethyl causes a 6-fold increase in activity<br />Reducing...
Glucocorticoid Inhibitors<br />Generated using a 30% cutoff<br />
Glucocorticoid Inhibitors<br />Suggests that electrondensity is also important<br />Lower π density possibly correlates to...
Glucocorticoid Inhibitors<br />These observations match those made by Takahashi et al.<br />More detailed graphs exhibit l...
SALI Graphs & Predictive Models<br />The graph view allows us to view SAR’s and identify trends easily<br />The aim of a Q...
Measuring Model Quality<br />A QSAR model should easily encode the “rolling hills”<br />A good model captures the most sig...
SALI Curves<br />
QSAR Model Comparisons<br />QSAR has traditionally used statistical or machine learning methods<br />But ‘Q’ is for quanti...
QSAR Model Comparisons<br />Derived from a pharmacophore model<br />Derived from a docking model<br />Holloway, K. et al, ...
Acknowledgements<br />John Van Drie<br />Gerry Maggiora<br />MicLajiness<br />JurgenBajorath<br />
Upcoming SlideShare
Loading in …5
×

A Network Visualization of Structure Activity Landscapes

1,274 views

Published on

0 Comments
2 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
1,274
On SlideShare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
20
Comments
0
Likes
2
Embeds 0
No embeds

No notes for slide

A Network Visualization of Structure Activity Landscapes

  1. 1. A Network Visualization of Structure Activity Landscapes<br />Rajarshi Guha<br />NIH Chemical Genomics Center<br />March 24, 2010<br />National ACS Meeting, San Francisco<br />
  2. 2. Structure Activity Relationships<br />Similar molecules will have similar activities<br />Small changes in structure will lead to small changes in activity<br />One implication is that SAR’s are additive<br />This is the basis for QSAR modeling<br />Martin, Y.C. et al., J. Med. Chem., 2002, 45, 4350–4358<br />
  3. 3. Exceptions Are Easy to Find<br />Ki = 39.0 nM<br />Ki = 1.8 nM<br />Ki = 10.0 nM<br />Ki = 1.0 nM<br />Tran, J.A. et al., Bioorg. Med. Chem. Lett., 2007, 15, 5166–5176<br />
  4. 4. Structure Activity Landscapes<br />Rugged gorges or rolling hills?<br />Small structural changes associated with large activity changes represent steep slopes in the landscape<br />But traditionally, QSAR assumes gentle slopes<br />Machine learning is not very good for special cases<br />Maggiora, G.M., J. Chem. Inf. Model., 2006, 46, 1535–1535<br />
  5. 5. Structure Activity Landscapes<br />
  6. 6. Characterizing the Landscape<br />A cliff can be numerically characterized<br />Structure Activity Landscape Index (SALI)<br />Cliffs are characterized by elements of the matrix with very large values<br />Guha, R.; Van Drie, J.H., J. Chem. Inf. Model., 2008, 48, 646–658<br />
  7. 7. Fingerprints<br />Lots of types of fingerprints<br />Indicates the presence or absence of a structural feature <br />Length can vary from 166 to 4096 bits or more<br />Fingerprints usually compared using the Tanimoto metric<br />
  8. 8. Visualizing the SALI Matrix<br />
  9. 9. Visualizing SALI Values<br />Alternatives?<br />A heatmap is an easy to understand visualization<br />Coupled with brushing, can be a handy tool<br />A more flexible approach is to consider a network view of the matrix<br />The SALI graph<br />Compounds are nodes<br />Nodes i,j are connected if SALI(i,j) > X<br />Only display connected nodes<br />
  10. 10. Visualizing the SALI Graph<br />Nodes are ordered such that the tail node in an edge has lower activity than the head node<br />
  11. 11. Varying the Cutoff<br />The cutoff controls the complexity of the graph <br />Higher cut offs will highlight the most significant activity cliffs<br />
  12. 12. Varying Fingerprint Methods<br />Shorter fingerprints will lead to more “similar” pairs<br />Requires a higher cutoff to focus on significantcliffs<br />
  13. 13. Varying the Similarity Metric<br />
  14. 14. Different Molecular Representations<br />The nature of the representation can significantlyaffect the landscape<br />SALI matrices for a benzodiazepine dataset, generated using a 2D and a 3D representation<br />Sutherland, J.J. et al., J. Chem. Inf. Comput. Sci., 2003, 43, 1906–1915<br />
  15. 15. Different Activity Representations<br />Using the Hill parameters from a dose-response curve represents richer data than a single IC50<br />
  16. 16. SALI Curves from DRCs<br />No difference in major cliffs<br />Some of the minor cliffs are highlighted using the DRC instead of IC50<br />IC50<br />DRC<br />
  17. 17. Better Visualization - SALIViewer<br />http://sali.rguha.net<br />
  18. 18. Glucocorticoid Inhibitors<br />62 dihydroquinolinederivatives<br />IC50’s reported, some values were censored<br />50% SALI graph generated using 1052 bit BCI fingerprints<br />Takahashi, H. et al, Bioorg. Med. Chem. Lett., 2007, 17, 5091–5095<br />
  19. 19. Glucocorticoid Inhibitors<br />62 dihydroquinolinederivatives<br />IC50’s reported, some values were censored<br />50% SALI graph generated using 1052 bit BCI fingerprints<br />Takahashi, H. et al, Bioorg. Med. Chem. Lett., 2007, 17, 5091–5095<br />
  20. 20. Glucocorticoid Inhibitors<br />Moving from ally or phenylethyl to ethyl causes a 6-fold increase in activity<br />Reducing bulk at this position seems to improve activity<br />But ethyl is not much smaller than allyl<br />We need more detail<br />07-20 2000 nM<br />07-23 2000 nM<br />07-17 355 nM<br />
  21. 21. Glucocorticoid Inhibitors<br />Generated using a 30% cutoff<br />
  22. 22. Glucocorticoid Inhibitors<br />Suggests that electrondensity is also important<br />Lower π density possibly correlates to increased activity<br />Confirmed by 07-23 -> 07-18<br />07-15 -> 07-17 is interesting since the change increases the bulk<br />07-20 2000 nM<br />07-15 2000 nM<br />07-18 710 nM<br />07-17 355 nM<br />
  23. 23. Glucocorticoid Inhibitors<br />These observations match those made by Takahashi et al.<br />More detailed graphs exhibit longer paths that focus on the bulk of side chains at the C4–α position<br />A number of paths consider changes to the epoxide substitution<br />Usually of length 1<br />Highlights the fact that bulk at the C4–α has greater impact on activity than epoxide substitutions<br />The SALI graph stresses the non-linearity of the SAR<br />
  24. 24. SALI Graphs & Predictive Models<br />The graph view allows us to view SAR’s and identify trends easily<br />The aim of a QSAR model is to encode SAR’s<br />Traditionally, we consider the quality of a model in terms of RMSE or R2<br />But in general, we’re not as interested in RMSE’s as we are in whether the model predicted something as more active than something else <br />What we want to have is the correct ordering<br />We assume the model is statistically significant<br />
  25. 25. Measuring Model Quality<br />A QSAR model should easily encode the “rolling hills”<br />A good model captures the most significantcliffs<br />Can be formalized as <br />How many of the edge orderings of a SALI graph does the model predict correctly?<br />Define S (X ), representing the number of edges correctly predicted for a SALI network at a threshold X<br />Repeat for varying X and obtain the SALI curve<br />
  26. 26. SALI Curves<br />
  27. 27. QSAR Model Comparisons<br />QSAR has traditionally used statistical or machine learning methods<br />But ‘Q’ is for quantitative – lots of ways to get a quantitative model<br />A model can encode a SAR in twoways<br />Implicit (via surrogates)<br />Explicit (via a physical model)<br />
  28. 28. QSAR Model Comparisons<br />Derived from a pharmacophore model<br />Derived from a docking model<br />Holloway, K. et al, J. Med. Chem., 1995, 38, 305–317<br />Cavalli, A. et al, J. Med. Chem., 2002, 45, 3844–3853<br />
  29. 29. Acknowledgements<br />John Van Drie<br />Gerry Maggiora<br />MicLajiness<br />JurgenBajorath<br />

×