Skip to content

Network Primer

A brief introduction to microbial co-occurrence networks, taught immediately after machine learning because the two methods answer complementary questions. ML asks "which features distinguish groups?"; networks ask "which features go together, regardless of group?"

What this lecture covers

  • What a co-occurrence network is and what each node and edge represent
  • Correlation-based vs compositional-aware approaches (SparCC, SPIEC-EASI, SCNIC)
  • How to read a network: hubs, modules, sign of correlation, edge weights
  • When networks tell you something machine learning misses, and when they mislead
  • Quick view of the workshop dataset's network for orientation

Hands-on follow-along

The full hands-on network build is not part of the in-person walkthrough, this block is a primer. Resources for building your own networks live on the QIIME2 forum's networking thread and in the SCNIC documentation.

Next: Longitudinal Analysis