Skip to content

Latest commit

 

History

History
69 lines (68 loc) · 5.72 KB

data-management.md

File metadata and controls

69 lines (68 loc) · 5.72 KB
Raw

Effective Practices for Data Management in Biomedical Research

Contributors (alphabetical): Paul Billing-Ross, Daniel Cotter, Tao Wang,

Terminology

  • Data generation site: An institution, such as a University or research lab that is responsible for generating processed experimental data and submitting it to a centralized data repository (Data Coordination Center)
  • Data generator: A person such as a technician, graduate student, post-doc, or other researcher who is responsible for transforming “raw data” (e.g. biological sample, sequencing reads) into an experiment-specific processed data (e.g. DNA variants, RNA-seq results, imaging data). This person is responsible for sharing meta/data with the data manager.
  • Data coordination center (DCC): The organization responsible for compiling data collected from every member data generation site and transforming it into a medium that is accessible to the broader research community
  • Data manager: A person with the technical expertise and responsibility to compile data generated at a data generation site and, …
    • Demonstrate to all stakeholders the value of effective data management
    • Establish effective practices for data generators to manage meta/data during the data generation and sharing process
    • Establish effective practices for data generators to share meta/data
    • Recommend technologies to improve the data generation & sharing process
    • If necessary, generate new tools to streamline procedures
    • Streamline the process of getting metadata from 3rd parties/vendors
    • Work with data generators to streamline data generation and sharing procedures
    • Ensure the completeness of data and metadata submission
    • Ensure the validity of the data and metadata
    • Perform necessary metadata cleaning and massaging
    • Restructure meta/data to fit the structure required by the DCC
  • Data sharing: The process of the data generator sharing the processed data with the data manager, in preparation for data submission
    • Also can involve moving data around internally, by data manager (outside of DCC)
  • Data submission: The process of the data manager submitting the compiled and finalized dataset to the data coordination center
    • Also can involve submission to internal LIMS system, likely before DCC submission
    • Data > Internal LIMS > DCC LIMS
    • Includes metadata submission as well as data submission (50/50)
    • Different modes of submission for different stakeholders
  • Project manager: A person who sends emails.

Goals

  • Eliminate points of failure between the data generator, data manager, and data coordination center (DCC)
  • (As a data manager) Make a significant effort to understand the workflows and challenges of researchers, and help develop solutions that make their lives easier
  • Standardize methods for managing data (across different projects)
    • Leverage existing infrastructure & tools between projects
    • Leverage existing knowledge
    • Minimize the cognitive load and migration load, to new methods
    • Limitations:
      • Constraints from DCC
      • Resistance to change
  • Federate data sharing to support different research populations
    • Researchers within a lab
    • Researchers with an institution
    • Researchers outside of the institution

Project initiation

  • Formalize the role of the data manager in project management.
    • Recognize that effective project management cannot occur without data management.
    • The data manager becomes a de facto project manager by virtue of their position at the center of all data operations. They should be formally recognized as a project manager or clearly establish a mode of working in tandem with the project manager
    • Establish the scope of responsibilities (for the data manager)
  • Have everyone go through checklist of prerequisites for the data generation and management process
  • Establish agreed upon modes of communication (Email, Slack, etc.) and what kind of information should be exchanged using them.
    • People use email/slack when things don’t go as planned
  • Establish how data management meetings will occur
    • For smaller groups this, data management discussions may occur in the context of other meetings (e.g lab meetings)
    • For large consortiums, with many members, it may be necessary to schedule separate meetings dedicated to data management
  • Establish what kind of project management software will be used (Trello, Jira, GitHub, Asana, etc.)
  • The data manager should try to establish a connection with at least one person within each data generation site
    • Confirm that established practices are being followed
    • Resolve interpersonal or technical issues that arise
  • Establish mapping of terms and IDs within project resources and between other projects
    • Identify ontologies to use or how terms relate to other ontologies
    • Map site specific IDs to global IDs or establish what/how IDs will be used

Data generation

  • Work with the DCC and researchers to establish the dictionary of metadata that will be included with data
  • Establish the data sharing methods and have all data generators run through the procedures with a dummy dataset to expose any road blocks (e.g. permissions issues)
  • Based on the frequency of submissions and volume of data, establish a schedule for data sharing that leaves enough buffer time to resolve any meta/data issues

Data sharing

  • Pre-hoc: If possible, use structured templates for data submission (e.g. web page, Google Forms)
    • Also for issue submission
      • Ensure completeness of issue tickets (e.g. code snippet, SCG submission scripts)
    • Set expectations for communication
  • Post-hoc: If a pre-hoc solutions is not feasible or adequate, write a program that will validate the structure and format of data once it has been submitted to the data manager