MATISSE Modular Analysis for Topology of Interactions and Similarity SEts

1
MATISSE - Modular Analysis for Topology of
Interactions and Similarity SEts

• Igor Ulitsky and Ron Shamir
• School of Computer Science
• Tel Aviv University

2
Motivation

• Detect functional modules groups of
• interacting proteins
• co-expressed genes
• Integrative analysis - can identify weaker
  signals
• Extraction of modules relevant to the profiled
  conditions

3
MATISSE

• Works with two graphs
• Interaction graph (unweighted)
• Edges PPI or PDI
• Similarity graph (weighted full)
• Heavy edges Similar expression patterns
• Identifies sets of genes that induce
• a heavy subnetwork of similarities, and
• a connected subnetwork of interactions

4
Front vs Back nodes

• Only variant genes (front nodes) have meaningful
  similarity values
• These can be linked by not regulated genes (back
  nodes).
• Back nodes correspond to
• Post-translational regulation
• Partially regulated pathways
• Unmeasured transcripts
• Molecular scaffolds

5
MATISSE workflow

• All the variations follow
• Seed generation
• Greedy optimization
• Significance filtering

6
Advantages of MATISSE

• No need for confidence estimation on individual
  measurements
• Works even when only a fraction of the genes
  expression patterns are informative
• Can handle any similarity data
• No need to prespecify the number of modules

7
Osmotic shock response of S. cerevisiae

• Constructed a network of 6,246 genes, 65,990
  protein-protein and protein-DNA interactions
• 133 experimental conditions response of
  perturbed strains to osmotic shock
• 2,000 genes selected as front nodes based on
  their variation

8
GO and promoter analysis
(c)
9
Pheromone response subnetwork
Back
Key TFs involved in pheromone response
Front
10
MATISSE in Expander

• First, the user needs to load a network
• Very simple SIF format currently supported
• MATISSE executed as any clustering algorithm
• Main parameters
• Beta (0-1) controls how similar should the
  expression profiles in each module be
• Maximal module size
• The set of genes that are used as front nodes
  (genes with variant expression profiles, should
  be lt3000 for a reasonable performance)

11
MATISSE in Expander
Front node
Back node