JSM 2017

Guiding Principles for Interactive Graphics Based on LIBD Data Science Projects

Transcript

1. 11
   Guiding Principles for Interactive
   Graphics Based on LIBD Data
   Science Projects
   Leonardo Collado-Torres
   @fellgernon
   https://speakerdeck.com/lcolladotor
   

   View Slide

2. CHALLENGES
   2
   Before you jump in the world of interactive graphics
   

   View Slide

3. CHALLENGES
   3
   Before you jump in the world of interactive graphics
   make sure that you are aware of some of
   the drawbacks
   

   View Slide

4. eQTL browser
   4 http://www.nature.com/nbt/journal/v35/n4/full/nbt.3838.html
   

   View Slide

5. CHALLENGES
   5
   D3: new language, new tools
   http://nvd3.org//examples/lineWithFocus.html
   https://github.com/ramnathv/rCharts
   

   View Slide

6. CHALLENGES
   6 https://github.com/nachocab/clickme
   https://github.com/lcolladotor/ballgownR-devel
   Interpretability
   

   View Slide

7. CHALLENGES
   7
   Data resolution
   https://github.com/nachocab/clickme
   https://github.com/lcolladotor/ballgownR-devel
   

   View Slide

8. CHALLENGES
   8
   

   View Slide

9. CHALLENGES
   9
   Dependencies
   https://github.com/nachocab/clickme
   https://github.com/lcolladotor/ballgownR-devel
   

   View Slide

10. CHALLENGES
    10
    rMaps case
    https://github.com/ramnathv/rMaps
    

    View Slide

11. CHALLENGES
    11
    Nowadays: no map!
    https://github.com/ramnathv/rMaps
    http://lcolladotor.github.io/
    

    View Slide

12. CHALLENGES
    12
    External dependencies
    Now missing:
    https://dl.dropboxusercontent.com/u/10794332/mx_states.json
    https://github.com/ramnathv/rMaps
    http://lcolladotor.github.io/
    

    View Slide

13. CHALLENGES
    13
    Quick summary
    • Interpretability
    • Data resolution: loading and
    sharing
    • Dependencies for deploying
    • External dependencies: less control
    

    View Slide

14. 29
    eQTL browser
    Scenario I: for publication
    • Bill Ulrich
    14
    

    View Slide

15. eQTL browser
    15 http://www.nature.com/nbt/journal/v35/n4/full/nbt.3838.html
    

    View Slide

16. eQTL browser
    Thousands of eQTLs
    16
    http://www.biorxiv.org/content/early/2017/04/05/124321
    

    View Slide

17. eQTL browser
    Search for a gene
    17 http://eqtl.brainseq.org/
    

    View Slide

18. eQTL browser
    Choose an eQTL to explore
    18 http://eqtl.brainseq.org/
    

    View Slide

19. eQTL browser
    View the data
    19 http://eqtl.brainseq.org/
    

    View Slide

20. eQTL browser
    View all the data + mouse-over display
    20 http://eqtl.brainseq.org/
    

    View Slide

21. eQTL browser
    Collection of data
    21 http://eqtl.brainseq.org/
    

    View Slide

22. eQTL browser
    Link to
    major
    community
    tools
    22 http://eqtl.brainseq.org/
    

    View Slide

23. eQTL browser
    Customized view with LIBD’s data
    23 http://eqtl.brainseq.org/
    

    View Slide

24. eQTL browser
    Scenario 1 summary
    24 http://eqtl.brainseq.org/
    • Re-use established tools as much as possible
    • Keep the information users will look at
    • Simple graphics
    • Speed is important
    • While developing: keep in mind users will
    want changes
    

    View Slide

25. Scenario II: for common use
    25
    • Stephen Semick
    shinycsv
    

    View Slide

26. Lots of tabular data: how can we make it easy to explore?
    26
    shinycsv
    

    View Slide

27. Excel’s recommended charts
    27
    shinycsv
    

    View Slide

28. Our solution: shinycsv
    28
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    

    View Slide

29. Simple summary statistics
    29
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    

    View Slide

30. Subset, search, re-order
    30
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    

    View Slide

31. Check new summary statistics after sub setting
    31
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    

    View Slide

32. Visualize a variable: just select it
    32
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    

    View Slide

33. Includes summary statistics
    33
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    

    View Slide

34. Plot changes by variable type
    34
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    

    View Slide

35. Plot two variables and change colors
    35
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    

    View Slide

36. Upload your data: handles different formats
    36
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    Download data if you edited it
    

    View Slide

37. Reproduce and learn from shinycsv
    37
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    

    View Slide

38. Our solution: shinycsv
    38
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv-showcase/
    

    View Slide

39. Using shinycsv non-interactively
    39
    shinycsv
    https://jhubiostatistics.shinyapps.io/shinycsv/
    Create PDFs with many plots, then explore manually
    

    View Slide

40. Scenario II summary
    40
    shinycsv
    http://lcolladotor.github.io/2017/01/20/Easily-explore-a-table-with-shinycsv
    • Re-use established tools as much as possible
    • Simple graphics: can be made prettier later
    • Minimize user options: try to make best
    guess
    • Might lead to unexpected use cases
    

    View Slide

41. Collaborators
    41
    Hopkins
    Jeff Leek
    Shannon Ellis
    Ben Langmead
    Chris Wilks
    Kai Kammers
    Kasper Hansen
    Margaret Taub
    OHSU
    Abhinav Nellore
    LIBD
    Andrew Jaffe
    Emily Burke
    Stephen Semick
    Carrie Wright
    Amanda Price
    Nina Rajpurohit
    Bill Ulrich
    

    View Slide

42. 11
    Guiding Principles for Interactive
    Graphics Based on LIBD Data
    Science Projects
    Leonardo Collado-Torres
    @fellgernon
    https://speakerdeck.com/lcolladotor
    

    View Slide