Skip to content

Latest commit

 

History

History
252 lines (213 loc) · 11.1 KB

project-management.md

File metadata and controls

252 lines (213 loc) · 11.1 KB

Project Management Tools

This is an overview of a variety of tools, platforms, and philosophies of managing research data in a FAIR context.

Reproducibility Benefits

The advantages of incorporating reproducible methods into a research management workflow are numerous, benefiting the individual researcher, the research team, and the broader scientific community. These advantages primarily center around minimizing risk, improving documentation, facilitating collaboration, and enhancing reputation.

Key advantages include:

1. Risk Mitigation and Error Prevention

Reproducible methods protect against loss of work and ensure the reliability of results:

  • Avoid Disaster: Reproducible practices help avoid disaster by ensuring that data and analyses are not lost ("this was my only copy!"). This is often tied to good Research Data Management (RDM), which includes recommendations, schemes, and standards for practices like storage and backup.
  • Ensuring Consistency: When analysis is reproducible (same data, same analysis), it ensures that the same results can be obtained consistently, validating the integrity of the computational steps. The use of standardized structures, such as the BIDS standard for neuroimaging data, contributes significantly to this consistency.

2. Improved Documentation and Reporting

Structured, reproducible workflows streamline the process of publishing and sharing findings:

  • Helps Writing Papers: Reproducible methods help writing papers. Tools and frameworks like Nipoppy, which manage configurations (config.json, descriptor.json, invocation.json), simplify the generation of detailed methodology sections.
  • Reviewer Trust: Reproducibility helps reviewers see it your way. Clear methods and documented steps allow reviewers and journal editors to verify the claims easily.
  • Supports Transparency (FAIR Principles): RDM and reproducible analyses make research FAIR (Findable, Accessible, Interoperable, and Reusable). Repositories and standardized formats ensure proper documentation, persistent identifiers, and licenses are applied to research outputs. Sharing derived results in an interoperable and understandable way makes them reusable and reproducible.

3. Enhanced Collaboration and Continuity

Reproducibility ensures that projects can be easily continued, extended, and utilized by others:

  • Enables Continuity: Reproducible methods enable continuity of the project. If a team member leaves, the next person can easily understand and pick up the workflow without ambiguity.
  • Facilitates Open Science and Data Sharing: Reproducible Data Management (RDM) greatly facilitates data sharing and Open Science. When code and data are placed in containers (like Docker or Singularity) and described via formal standards (like BIDS Apps API), other researchers can quickly adopt and run the analysis pipeline. Nipoppy, for instance, uses containerized software and descriptor files to manage the execution of pipelines, making them portable and reusable across different computing environments.
  • Avoids Duplication of Effort and Funding: Sharing reproducible work prevents other researchers from unnecessarily duplicating efforts, which is beneficial to the overall scientific ecosystem.

4. Professional and Career Benefits

  • Builds Reputation: Reproducible work helps to build your reputation in the scientific community.
  • Increases Citation Count: Open source work is more likely to be reused and cited, benefiting the author.

Dataset Hosting

Based on the sources, several resources are available to host and share neuroimaging datasets, ranging from domain-specific repositories to general-purpose platforms, all designed to support Research Data Management (RDM) and Open Science principles.

Dedicated Neuroimaging Data Repositories

These repositories are specifically recognized within the neuroimaging and neuroscience research community:

  • LORIS: This system was customized for data entry, storage, and dissemination of the PREVENT-AD program data.

    • The PREVENT-AD cohort uses two LORIS databases for data dissemination: one for Open Access (for basic demographics and longitudinal neuroimaging raw data, e.g., https://openpreventad.loris.ca) and one for Registered Access (for sensitive material like cognitive, medical, neurosensory, genetic, and CSF data, e.g., https://registeredpreventad.loris.ca).
    • The PREVENT-AD repositories are also discoverable via the Canadian Open Neuroscience Platform (CONP) interface (https://portal.conp.ca).

  • OpenNeuro: This platform is listed as a repository for data sharing.

    • It is considered very popular for neuroimaging data, especially data following the Brain Imaging Data Structure (BIDS) standard.
    • It is designed for publishing certain neuroimaging modalities, uses the BIDS standard, and supports version control. However, it generally does not offer access control, though there may be an embargo period.

  • Brainlife: This platform is available for work and publishing of neuroimaging data.

    • It supports the BIDS standard and offers access control and version control.

  • OMEGA (Open MEG Archive): This repository is specifically mentioned as the place where the raw MEG and T1-weighted MRI scans from PREVENT-AD participants are available.

  • NITRC (Neuroimaging Tools and Resources Clearinghouse): This repository publishes neuroimaging data, tools, and other resources.

    • It already hosts popular datasets such as ABIDE, OASIS, and ADHD_200.
    • It offers access control and provides a file preview function.

  • DANDI: This repository publishes physiology data and supports the BIDS standard, access control, and version control.

  • NeuroVault, TemplateFlow, and EBRAINS: These platforms are listed as options for sharing derived data and results.

General-Purpose Data Repositories

These platforms can host various types of research data, including neuroimaging data derivatives:

  • Zenodo: This platform is a repository used for publishing any kind of data.

    • It is often utilized in the context of open science for uploading and installing community-developed pipelines (like those used in Nipoppy).
    • It is considered intuitive to navigate, offers access control, and supports version control, although it has storage limits.

  • OSF (Open Science Framework): This tool promotes open, centralized workflows throughout the research lifecycle, including storing and analyzing collected data.

    • It is suitable for work and publishing of any data type.
    • It provides a unique, persistent uniform resource locator (URL) or digital object identifiers (DOIs) and archival resource keys (ARKs).
    • It features built-in version control and allows third-party add-ons or integrations.

  • Figshare: This repository is used for publishing any type of data.

    • It is noted for being fast in up- and download, which is good for handling big data in large chunks.

  • NDA (National Database for Autism Research) (Implied): Listed as a repository for publishing any human data.

  • GIN: This tool is used for both work and publishing and can handle any kind of data.

Considerations for Choosing a Repository

When selecting a repository, researchers should ideally submit data to a domain or discipline-specific, community-recognized repository. Discipline-specific repositories often have more functionalities tailored for the type of data being shared. Repositories ensure long-term archiving and preservation, which differentiates them from collaborative development tools like GitHub or GitLab. Repositories promote the FAIR Principles (Findable, Accessible, Interoperable, Reusable) by assigning persistent identifiers, using metadata standards (like BIDS), providing licenses, and determining access levels.

Project Frameworks

The sources describe several types of frameworks that exist for research project management, ranging from high-level guiding principles and documentation standards to comprehensive data and pipeline management systems, particularly within the field of neuroinformatics.

Research Data Management (RDM) Frameworks

Research Data Management (RDM) provides the fundamental recommendations, schemes, and standards for practices of storing and annotating research data. RDM supports adherence to the Research Data Lifecycle. Key formal documentation frameworks include:

  • Research Data Management Plan (RDMP): This is a formal, living document that outlines the handling of research data both during and after the project. It functions as a roadmap for efficiently and securely managing data. RDMPs often detail the following aspects of a project:
    • Types of data (source, format, volume).
    • Contextual details (metadata and documentation).
    • Storage, backup, and security procedures.
    • Provisions for protection/privacy.
    • Policies for re-use by other researchers.
    • Access and sharing plans.
    • Archiving and providing access (long-term preservation plans).
    • Roles and plan oversight (who is responsible for implementation).
  • Software Management Plan (SMP): A formal plan that describes how a specific software project will be developed and maintained, including version control, licensing, and license compatibility. This plan details how the software will be made available to users and how its contribution to research will be measured.
  • FAIR Guiding Principles: RDM makes research FAIR—Findable, Accessible, Interoperable, and Reusable. Repositories support these principles by assigning persistent identifiers, using metadata standards, and providing licenses.

Neuroimaging Workflow and Data Management Frameworks

These tools facilitate open science and structured research collaboration across the entire project lifecycle:

  • Open Science Framework (OSF): OSF is a tool that promotes open, centralized workflows and captures various aspects of the research lifecycle, including developing ideas, designing studies, storing/analyzing collected data, and writing/publishing reports or preregistrations. It features built-in version control and persistent URLs/DOIs for research objects.
  • Project Organization Templates: Frameworks that provide reusable templates for structuring project folders and files, such as:
    • Cookiecutter.
    • TONIC Research Folder Structure.

Version Control Tools

These tools are essential components of reproducibility and project management frameworks:

  • Code and Data Versioning: Tools mentioned for version controlling code, data, and text include Git, SVN, Git LFS (Git Large File Storage), git-annex, DataLad, and GIN.
  • Containerization: Tools like Docker and Singularity/Apptainer are used to containerize software, which is crucial for reproducible pipeline deployment within frameworks like Nipoppy.