Skip to main content

Find data across the network

In FLNet, discovery is designed to help you identify relevant datasets without assuming that raw data is centrally pooled or immediately accessible.

What discovery usually means​

When you search the network, you are typically exploring:

  • which sites have relevant data
  • how that data is structured
  • which cohorts or variables appear to be available
  • what level of access or participation may be possible

This is a metadata-driven process first, not an unrestricted data export flow.

Typical discovery workflow​

1. Start from the scientific question​

Before searching, define:

  • what population you need
  • which variables matter
  • whether you need one site or several
  • whether the analysis is descriptive, predictive, or federated

Discovery works much better when the question is specific.

2. Search and filter available metadata​

Use the platform to narrow down datasets by:

  • domain or disease area
  • available variables
  • cohort characteristics
  • site participation

At this stage, you are trying to answer: Which sites are likely relevant?

3. Review quality and fit​

Once you find candidate datasets, review:

  • whether the schema matches your expected inputs
  • whether enough sites appear relevant
  • whether the data standard aligns with your planned tool or workflow

4. Decide what kind of project you need​

After discovery, the next step is usually one of these:

  • create a centralized analysis project on data you are allowed to upload or access
  • create a federated analysis project spanning multiple sites
  • contact the relevant data holders for further governance or approval steps

Important expectation: discovery is not permission​

Seeing that data exists does not automatically mean you can run anything against it.

A discovered dataset may still require:

  • project-level approval
  • client-side access configuration
  • specific tool support
  • additional coordination with participating sites

This is a feature of the system, not a limitation. It keeps discovery useful without eroding local control.

Good practice when evaluating candidate data​

Ask these questions early:

  • Is the question answerable with metadata-guided discovery alone?
  • Do I need local variables to be harmonized first?
  • Is the analysis tool compatible with the expected input structure?
  • Will this be a centralized or federated workflow?

Next step​

Once you have identified relevant datasets, continue with Use found data for federated data analysis.