Exercise 1: Co‑expression networks

Prof. Patrick E. Meyer

Load the file dream4-multifact4.Rdata, containing an expression dataset (called dataset) and a corresponding gold‑standard network (called true.net) located in the public/biol0021/ directory (also available from the course webpage).

  1. Transform the dataset into a co‑expression network (using squared Spearman’s correlation).
  2. Compute the PR‑curves for our co‑expression network (given the provided gold‑standard and the minet package).
  3. Check if using the CLR algorithm to eliminate indirect arcs improves the PR‑curves of our co‑expression network.
  4. Transform the CLR weighted adjacency matrix into a graph by keeping only the links with a score ≥ 0.9 (hint: transform the 0‑1 adjacency matrix into an igraph object).
  5. Give the size (number of edges), the density and the diameter of the graph.
  6. How many communities of genes (and of what size) – i.e., groups of genes that likely share functional goals – do you detect in that network?