This comes down to the difference between pathways and networks.
- schematic of biological knowledge
- indicates many different types of interactions
- tend to be human curated, hence tend to cover only a fraction of the genome
- eg. KEGG, Reactome
- a set of bimolecular interactions represented as a graph (directed or undirected)
- can capture well-studied and not so well-studied biology
- much higher coverage than pathways, but supported by less data
- eg. BioGRID, IntAct
Pathway analysis uses more complex models to try to utilize the additional mechanistic and regulatory information that pathways contain. Network analysis usually means overlaying a set of genes on top of a pre-existing network and asking whether they cluster in some region.
I would add this slightly nuanced definition:
Pathway analysis is analyzing a list of genes against an a priori set of pathways with known interactions for various systems, and identifying which pathway is most impacted.
Network analysis is analyzing how a set of genes are connected given a known set of direct interactions, and identifying which genes cluster in a network.