$30 off During Our Annual Pro Sale. View Details »

Atlas User Interfaces to Accelerate Data Access

Atlas User Interfaces to Accelerate Data Access

How can we design user interfaces that make data access effective and efficient? Introduces http://vitessce.io!

Nils Gehlenborg

June 28, 2022
Tweet

More Decks by Nils Gehlenborg

Other Decks in Science

Transcript

  1. Atlas User Interfaces to
    Accelerate Data Access
    @ngehlenborg ∙ http://gehlenborglab.org
    Human Cell Atlas General Meeting 2022
    Nils Gehlenborg, PhD
    Harvard Medical School

    View Slide

  2. Background
    Insights from Data Coordination Center projects
    NIH 4D Nucleome Consortium (Phase 1 & 2): 4DN DCIC

    PI: Park; since 2015
    NIH Human BioMolecular Atlas Program (Phase 1 & 2): HuBMAP HIVE

    PIs: Blood, Silverstein, Börner, Bar-Joseph, Gehlenborg; since 2018
    NIH Cellular Senescence Network: SenNet CODCC

    MPIs: Blood, Silverstein, Bar-Joseph; since 2021
    NIH Kidney Precision Medicine Project (Phase 2): KPMP KMAP

    MPIs: Krenzler, Himmelfarb; starting 2022

    View Slide

  3. How can we design
    user interfaces that make
    data access effective and efficient?

    View Slide

  4. How can we design
    user interfaces that make
    data access effective and efficient?

    View Slide

  5. How can we design
    user interfaces that make
    data access effective and efficient?
    deliver insight and minimize effort?

    View Slide

  6. How can we design
    user interfaces that make
    data access effective and efficient?
    do science and do it quickly?

    View Slide

  7. What are we doing in HuBMAP to
    accelerate data access?

    View Slide

  8. Data Knowledge
    HuBMAP Data Portal Human Reference Atlas
    - Data Ingestion and Curation
    - Uniform Processing Pipelines
    - Molecular Data Visualization
    - …
    - Common Coordinate Framework
    - Spatial Information about Samples
    - ASCT+B Tables
    - …
    HuBMAP View of the Cell Atlas World

    View Slide

  9. What cell types are where in the body?
    How are they spatially organized?
    What information was used to establish this?
    How reliable is this information?
    Where is the data behind this?

    View Slide

  10. Data Knowledge
    HuBMAP Data Portal Human Reference Atlas
    - Data Ingestion and Curation

    - Uniform Processing Pipelines

    - Molecular Data Visualization

    - …
    - Common Coordinate Framework

    - Spatial Information about Samples

    - ASCT+B Tables

    - …
    HuBMAP View of the Cell Atlas World

    View Slide

  11. Deliver Insight and Minimize Effort
    Do science and do it quickly
    Deliver Insight
    - Provide data visualization tools

    - Enable queries across datasets (work in progress)

    - Integrate data analysis tools (work in progress)
    Minimize Effort
    - Maximize browser-based portal capabilities, minimize clicks to action

    View Slide

  12. Data Visualization Tools
    Deliver Insight

    View Slide

  13. View Slide

  14. Data Visualization Tools
    Deliver Insight

    View Slide

  15. Data Visualization Tools
    Deliver Insight

    View Slide

  16. Data Visualization Tools
    Deliver Insight

    View Slide

  17. Data Visualization Tools
    Deliver Insight

    View Slide

  18. Data Visualization Tools
    Deliver Insight

    View Slide

  19. Data Visualization Tools
    Deliver Insight
    MS Keller et al., 2021

    doi:10.31219/osf.io/y8thv
    Vitessce

    http://vitessce.io

    View Slide

  20. Data Visualization Tools
    Single Dataset Visualization
    Deliver Insight

    View Slide

  21. Data Visualization Tools
    Single Dataset Visualization

    Clustering, segmentation, cell type assignments, etc. for single assay for single
    sample but multiple analysis outputs
    Deliver Insight

    View Slide

  22. Data Visualization Tools
    Single Dataset Visualization

    Clustering, segmentation, cell type assignments, etc. for single assay for single
    sample but multiple analysis outputs
    Multi Dataset Visualization

    Multiple samples of single assay, multiple assays for single sample, etc.
    Altas Visualization

    UIs to test hypotheses, build models, and make predictions
    Deliver Insight

    View Slide

  23. Queries Across Datasets
    Deliver Insight
    TS Liaw & N Gehlenborg, 2022

    Proceedings of the Design
    Society, 2, 221-230.
    doi:10.1017/pds.2022.24

    View Slide

  24. Queries Across Datasets
    Deliver Insight

    View Slide

  25. Queries Across Datasets
    Deliver Insight

    View Slide

  26. Queries Across Datasets
    Deliver Insight

    View Slide

  27. Queries Across Datasets
    Deliver Insight

    View Slide

  28. Queries Across Datasets
    Deliver Insight

    View Slide

  29. Queries Across Datasets
    Deliver Insight

    View Slide

  30. Queries Across Datasets
    Deliver Insight

    View Slide

  31. Queries Across Datasets
    Deliver Insight

    View Slide

  32. Queries Across Datasets
    Deliver Insight

    View Slide

  33. Queries Across Datasets
    Deliver Insight

    View Slide

  34. Queries Across Datasets
    Deliver Insight

    View Slide

  35. Queries Across Datasets
    Deliver Insight

    View Slide

  36. Data Analysis Tools
    Deliver Insight
    Jupyter Notebooks and workspaces in HuBMAP Data Portal
    Focus on computational analysts
    - Support cutting edge work
    - Extensible solution that enables community contributions
    - Combine HuBMAP data with user data
    - Use HuBMAP infrastructure for computation and storage
    - Enable pick up where portal UI functionality leaves off

    View Slide

  37. Data Analysis Tools
    Deliver Insight
    Today
    Create and download Jupyter
    Notebooks to execute locally
    or in Google Colab

    View Slide

  38. Data Analysis Tools
    Deliver Insight
    Today
    Create and download Jupyter
    Notebooks to execute locally
    or in Google Colab

    View Slide

  39. Data Analysis Tools
    Deliver Insight
    Today
    Create and download Jupyter
    Notebooks to execute locally
    or in Google Colab

    View Slide

  40. Data Analysis Tools
    Deliver Insight
    Today
    Create and download Jupyter
    Notebooks to execute locally
    or in Google Colab

    View Slide

  41. View Slide

  42. View Slide

  43. View Slide

  44. Data Analysis Tools
    Deliver Insight
    Today
    Create and download Jupyter
    Notebooks to execute locally
    or in Google Colab

    View Slide

  45. Data Analysis Tools
    Deliver Insight
    Today
    Create and download Jupyter
    Notebooks to execute locally
    or in Google Colab

    View Slide

  46. Data Analysis Tools
    Deliver Insight
    Today
    Create and download Jupyter
    Notebooks to execute locally
    or in Google Colab
    Near Future
    Create and execute Jupyter
    Notebooks in Jupyter Lab in
    HuBMAP infrastructure

    View Slide

  47. Data Analysis Tools
    Deliver Insight
    Today
    Create and download Jupyter
    Notebooks to execute locally
    or in Google Colab
    Near Future
    Create and execute Jupyter
    Notebooks in Jupyter Lab in
    HuBMAP infrastructure

    View Slide

  48. Deliver Insight and Minimize Effort
    Do science and do it quickly
    Deliver Insight
    - Provide data visualization tools

    - Enable queries across datasets (work in progress)

    - Integrate data analysis tools (work in progress)

    Minimize Effort
    - Maximize browser-based portal capabilities, minimize clicks to action

    View Slide

  49. Deliver Insight and Minimize Effort
    Do science and do it quickly
    Deliver Insight
    - Provide data visualization tools

    - Enable queries across datasets (work in progress)

    - Integrate data analysis tools (work in progress)

    Minimize Effort
    - Maximize browser-based portal capabilities, minimize clicks to action
    Deliver Insight
    - Provide data visualization tools

    - Enable queries across datasets (work in progress)

    - Integrate data analysis tools (work in progress)

    Minimize Effort
    - Maximize browser-based portal capabilities, minimize clicks to action

    - Optimize task-based creation of data downloads (work in progress)

    View Slide

  50. Task-Based Creation of Data Downloads
    Minimize Effort
    Interactive Mockup

    View Slide

  51. View Slide

  52. Task-Based Creation of Data Downloads
    Minimize Effort
    Interactive Mockup

    View Slide

  53. What are some takeaways of
    this work?

    View Slide

  54. Deliver Insight and Minimize Effort
    Contributor Interdependencies
    UI designers and developers are high value customers of any atlas data
    management and analysis infrastructure developed.
    UI designers and developers must deliver use cases that drive infrastructure
    development.
    UI
    Infra-
    structure
    User interfaces must not be an afterthought and coordination is critical!

    View Slide

  55. Deliver Insight and Minimize Effort
    Contributor Interdependencies
    UI
    Infra-
    structure
    Data
    Analysis
    Tools
    Data
    Vis
    Tools
    UI designers and developers are high value customers of any atlas data
    management and analysis infrastructure developed.
    UI designers and developers must deliver use cases that drive infrastructure
    development.
    User interfaces must not be an afterthought and coordination is critical!

    View Slide

  56. Deliver Insight and Minimize Effort
    Contributor Interdependencies
    UI
    Infra-
    structure
    Data
    Analysis
    Tools
    Data
    Vis
    Tools

    View Slide

  57. Deliver Insight and Minimize Effort
    Charting a Roadmap for Atlas User Interfaces
    Tier 1: View Data

    UIs to understand what questions can be asked
    Tier 2: Explore Data

    UIs to generate hypotheses
    Tier 3: Analyze Data

    UIs to test hypotheses, build models, and make predictions
    Cell atlas UIs are typically somewhere between Tier 1 and Tier 2 today.

    View Slide

  58. Challenges
    The Wild West of Cell Atlases
    Sociological
    - Diverse user audiences with different goals: user research is hard

    - Diverse teams contribute expertise to atlas building: coordination is hard
    Technological
    - Diverse data schemas, formats, and tools in use: integration is hard
    Scientific
    - Diverse analysis questions: science is not always easy

    View Slide

  59. Opportunities
    Calling up the Community
    Collaborate on shared challenges in atlas user interface development
    - Share use cases, personas, and other user research artifacts
    - Agree on common data schemas and formats
    Consider APIs user interfaces
    - Many analysts will interact with data primarily through APIs and serving
    their needs for efficient data access will accelerate science
    What else can we do?
    Postdoctoral Fellow
    (Senior) UI/UX Developer

    (Senior) Software Developer
    Apply for a
    position in
    my lab as a
    Email [email protected] if interested!

    View Slide