R Not Only in Production

I will share what our team has learned from successfully integrating R in all areas of our company's operations. InsightRX is a precision medicine company whose goal is to ensure that each patient receives the right drug at the optimal dose. At InsightRX, R is a first-class language that's used for purposes ranging from customer-facing products to internal data infrastructure, new product prototypes, and regulatory reporting. Using R in this way has given us the opportunity to forge fruitful collaborations with other teams in which we can both learn and teach.

Transcript

1. Kara Woo
   
   
   posit::conf(2023)
   R not only in production
   

   View Slide

2. View Slide

3. View Slide

4. View Slide

5. How can the strengths of R, and the
   people who use it, influence the
   organization as a whole?
   

   View Slide

6. Meet Angela
   

   View Slide

7. View Slide

8. Vancomycin can treat MRSA
   

   View Slide

9. View Slide

10. There’s no one dose that works for everyone
    

    View Slide

11. We use pharmacological models +
    patient’s unique characteristics to
    recommend the optimal dose
    

    View Slide

12. View Slide

13. Patient data, lab
    
    
    results, past doses
    

    View Slide

14. Patient data, lab
    
    
    results, past doses
    Recommended dose
    

    View Slide

15. R is core to our product
    

    View Slide

16. R is core to fulfilling our mission
    

    View Slide

17. View Slide

18. ETL
    

    View Slide

19. View Slide

20. View Slide

21. View Slide

22. View Slide

23. View Slide

24. • REST API
    
    
    • Data processing pipeline
    
    
    • Analyses and new models
    
    
    • Model validation
    
    
    • Regulatory reporting
    
    
    • New product prototypes
    
    
    • …
    

    View Slide

25. View Slide

26. View Slide

27. {
    "firstName": "First name",
    "lastName": "Last name",
    "dob": "Date of birth",
    "changeMyPassword": "Change my password",
    "logout": "Logout",
    ...
    }
    {
    "firstName": "Vorname",
    "lastName": "Nachname",
    "dob": "Geburtsdatum",
    "changeMyPassword": "Ändere mein Passwort",
    "logout": "Ausloggen",
    ...
    }
    

    View Slide

28. {
    "firstName": "First name",
    "lastName": "Last name",
    "dob": "Date of birth",
    "changeMyPassword": "Change my password",
    "logout": "Logout",
    ...
    }
    

    View Slide

29. {
    "firstName": "First name",
    "lastName": "Last name",
    "dob": "Date of birth",
    "age": "Age",
    "changeMyPassword": "Change my password",
    "logout": "Logout",
    ...
    }
    👀
    

    View Slide

30. {
    "firstName": "First name",
    "lastName": "Last name",
    "dob": "Date of birth",
    "age": "Age",
    "changeMyPassword": "Change my password",
    "logout": "Logout",
    ...
    }
    {
    "firstName": "Vorname",
    "lastName": "Nachname",
    "dob": "Geburtsdatum",
    "changeMyPassword": "Ändere mein Passwort",
    "logout": "Ausloggen",
    ...
    }
    👀
    

    View Slide

31. {
    "firstName": "First name",
    "lastName": "Last name",
    "dob": "Date of birth",
    "age": "Age",
    "changeMyPassword": "Change my password",
    "logout": "Logout",
    ...
    }
    {
    "firstName": "Vorname",
    "lastName": "Nachname",
    "dob": "Geburtsdatum",
    "age": "Alter",
    "changeMyPassword": "Ändere mein Passwort",
    "logout": "Ausloggen",
    ...
    }
    👀
    

    View Slide

32. The mindset of R as a first class
    language empowers us to solve
    problems.
    

    View Slide

33. View Slide

34. View Slide

35. View Slide

36. We can do this in R!
    

    View Slide

37. 💡
    
    
    Idea
    

    View Slide

38. 💡
    
    
    Idea
    “It works
    
    
    on my
    
    
    machine”
    

    View Slide

39. 💡
    
    
    Idea
    “It works
    
    
    on my
    
    
    machine”
    ? Final
    
    
    product
    ?
    ?
    

    View Slide

40. 💡
    
    
    Idea
    “It works
    
    
    on my
    
    
    machine”
    ? Final
    
    
    product
    ?
    ?
    

    View Slide

41. Reliability
    Deployability
    

    View Slide

42. We need to detect
    kidney injury accurately,
    all the time.
    

    View Slide

43. View Slide

44. Elevated creatinine suggests kidney damage
    Easy, right?
    • Data gaps
    
    
    • Out of order data
    
    
    • Data in di
    ff
    erent units
    
    
    • Missing baseline
    
    
    • Non-numeric data
    
    
    • Data entry errors
    

    View Slide

45. test_that("AKI stage is calculated correctly", {
    result <- calc_aki_stage(
    scr = c(0.7, 1.0),
    t = c(0, 24),
    egfr = c(60, 40)
    )
    expect_equal(result$stage, "stage 1")
    })
    

    View Slide

46. test_that("AKI stage is calculated correctly", {
    result <- calc_aki_stage(
    scr = c(0.7, 1.0),
    t = c(0, 24),
    egfr = c(60, 40)
    )
    expect_equal(result$stage, "stage 1")
    })
    

    View Slide

47. test_that("AKI stage is calculated correctly", {
    result <- calc_aki_stage(
    scr = c(0.7, 1.0),
    t = c(0, 24),
    egfr = c(60, 40)
    )
    expect_equal(result$stage, "stage 1")
    })
    

    View Slide

48. test_that("AKI stage is calculated correctly", {
    result <- calc_aki_stage(
    scr = c(0.7, 1.0),
    t = c(0, 24),
    egfr = c(60, 40)
    )
    expect_equal(result$stage, "stage 1")
    })
    

    View Slide

49. test_that("stage correct: 0.3 mg/dl rise", {
    # ...
    })
    test_that("stage correct: 1.5x baseline rise", {
    # ...
    })
    test_that("sorts times correctly", {
    # ...
    })
    test_that("handles non-numeric inputs", {
    # ...
    })
    test_that("handles only 1 observation", {
    # ...
    })
    

    View Slide

50. Jenny Bryan
    “If you use software that lacks automated
    tests, you are the tests.”
    

    View Slide

51. Get latest
    
    
    code
    Build the
    
    
    R package
    Run the
    
    
    tests
    Continuous integration
    

    View Slide

52. 5d01863 Add function to detect AKI ✅
    a28vf11 Rewrite AKI function to improve
    performance
    ❌
    682hcKa Fix handling of missing data ✅
    

    View Slide

53. Continuous integration
    • Find bugs quickly
    
    
    • Be lazy free up time for more important work
    
    
    • Make sure the code continues to behave as intended
    

    View Slide

54. 💡
    
    
    Idea
    “It works
    
    
    on my
    
    
    machine”
    ? Final
    
    
    product
    Testing
    Continuous
    
    
    Integration
    

    View Slide

55. 💡
    
    
    Idea
    “It works
    
    
    on my
    
    
    machine”
    Docker Final
    
    
    product
    Testing
    Continuous
    
    
    Integration
    

    View Slide

56. R
    
    
    R packages
    
    
    your code
    Docker
    

    View Slide

57. FROM rstudio/plumber:v1.1.0
    RUN Rscript -e "install.packages('clinPK')"
    COPY . /myproject
    WORKDIR /myproject
    ENTRYPOINT ["Rscript", "entrypoint.R"]
    Dockerfile
    

    View Slide

58. FROM rstudio/plumber:v1.1.0
    RUN Rscript -e "install.packages('clinPK')"
    COPY . /myproject
    WORKDIR /myproject
    ENTRYPOINT ["Rscript", "entrypoint.R"]
    Dockerfile
    

    View Slide

59. FROM rstudio/plumber:v1.1.0
    RUN Rscript -e "install.packages('clinPK')"
    COPY . /myproject
    WORKDIR /myproject
    ENTRYPOINT ["Rscript", "entrypoint.R"]
    Dockerfile
    

    View Slide

60. FROM rstudio/plumber:v1.1.0
    RUN Rscript -e "install.packages('clinPK')"
    COPY . /myproject
    WORKDIR /myproject
    ENTRYPOINT ["Rscript", "entrypoint.R"]
    Dockerfile
    

    View Slide

61. FROM rstudio/plumber:v1.1.0
    RUN Rscript -e "install.packages('clinPK')"
    COPY . /myproject
    WORKDIR /myproject
    ENTRYPOINT ["Rscript", "entrypoint.R"]
    Dockerfile
    

    View Slide

62. FROM rstudio/plumber:v1.1.0
    RUN Rscript -e "install.packages('clinPK')"
    COPY . /myproject
    WORKDIR /myproject
    ENTRYPOINT ["Rscript", "entrypoint.R"]
    Dockerfile
    https://github.com/karawoo/dockerdemo
    

    View Slide

63. Image: Royyan Razka, Noun Project
    

    View Slide

64. Get latest
    
    
    code
    Build the
    

    R package
    Run the
    
    
    tests
    Continuous integration
    

    View Slide

65. Get latest
    
    
    code
    Build the
    
    
    docker image
    Run the
    
    
    tests
    Continuous integration
    

    View Slide

66. 💡
    
    
    Idea
    “It works
    
    
    on my
    
    
    machine”
    Docker Final
    
    
    product
    Testing
    Continuous
    
    
    Integration
    

    View Slide

67. View Slide

68. View Slide

69. We are motivated to answer
    questions
    

    View Slide

70. View Slide

71. Database query was too slow
    

    View Slide

72. Database query was too slow
    
    
    Added an index to the database
    

    View Slide

73. Database query was too slow
    
    
    Added an index to the database
    
    
    Data processing pipeline was too slow
    

    View Slide

74. Database query was too slow
    
    
    Added an index to the database
    
    
    Data processing pipeline was too slow
    
    
    Removed the index
    

    View Slide

75. View Slide

76. If you can’t make changes because
    you’re afraid of breaking something,
    it’s already broken.
    

    View Slide

77. View Slide

78. We can visualize and
    communicate our findings
    

    View Slide

79. View Slide

80. Tests
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    

    View Slide

81. Tests
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    test_that("...", {
    # ...
    })
    become data
    

    View Slide

82. View Slide

83. We are used to dealing with
    uncertainty
    

    View Slide

84. View Slide

85. View Slide

86. Motivated to answer questions
    
    
    Skilled at visualizing and communicating results
    
    
    Able to handle uncertainty
    

    View Slide

87. View Slide

88. Using R throughout the stack =
    doing science throughout the stack
    

    View Slide

89. View Slide

90. pos.it/slido-keynote
    

    View Slide