2. Examples
Nature Reviews Genetics 15, 107–120 (2014) doi:10.1038/nrg3643
FBA = flux-balance analysis
• Topological enrichment can give broad overview
of impacted genes, proteins and metabolites
• Changes in biochemical domains corroborated
by multi-Omic data sets can be used to identify
robust candidates responsible for phenotypic
variation between comparisons
• Gene-gene, protein-protein or gene-protein
interaction networks can be used to
deconvolute ambiguous metabolic pathways
4. Biochemical Domain
Enrichment Analysis
• Genes/Proteins DAVID, AmiGo, etc GO:terms
• Genes/Proteins + Metabolites IMPaLA: Integrated Molecular
Pathway Level Analysis (http://impala.molgen.mpg.de/) pathways
1. Classify all species domains (e.g. biological process, pathway, etc)
2. Calculate probability of observing changes in species by chance
5. IMPaLA: Gene + Metabolite
pathway enrichment
Challenges:
•Removal of redundant information
•Preference of specific vs. generic pathways
•Visualization of gene + metabolite + pathway relationships
6. Determining significance of the
enrichment: Hypergeometric Test
How to calculate statistics to determine enrichment?
hit.num = 51 # number of significantly changed pathway
metabolites
set.num = 1455 # number of metabolites in pathway
full = 3358 # all possible metabolites in organism
q.size = 72 # number of significantly changed metabolites
phyper(hit.num-1, set.num, full-set.num, q.size, lower.tail=F)
= 1.717553e-06
7. GO Enrichment analysis:
Hierarchy of Redundancy (parents)
• GO is an ontology wherein enrichment is often
shared by children and parents.
• Difficult to co-visualize term hierarchy and gene to
term mapping
8. Enrichment networks:
Removing the Hierarchy of
Redundancy
Workflow:
1. If two nodes share all genes, drop least
enriched (highest p-value)
2. Filter terms based on enrichment
3. Display term to gene/protein
relationships as edges in a network
4. Map direction of change in
genes/proteins to network node
attributes
9. Enrichment Network
Mapping of parents through children
GO enrichment network displays:
• gene names associated with
each overrepresented term
• Fold change in protein
expression between two
groups (can be extended k>2
groups)
• Can display enrichment p-
value for each term
• Can incorporate metabolites
as children of genes
10. Empirical Networks
• Correlation based networks (CN)
(simple, tendency to hairball)
• GGM or partial correlation based
networks (advanced, preference
of direct over indirect
relationships
• *Increase in robustness with
sample size
10.1007/978-1-4614-1689-0_17