Reactome and WikiPathways both have apps for Cytoscape that pull their content into a graph model:

• ReactomeFI plugin
• WikiPathways app

In the case of the WikiPathways app, you have the explicit option of importing a pathway as "a pathway" (meaning a complete, annotated diagram) or as "a network" (meaning a simplified graph diagram). The network version is ideal for graph analytics, merging and extension.

The app exposes a REST API via Cytoscape as well, so you can programmatically import hundreds of (or all) pathways via an R or Python script. See RCy3 and py4cytoscape packages.

By the way, all the human pathways from Reactome are also hosted by WikiPathways. So, using the WikiPathways app, you can also pull in Reactome pathways (for human) into Cytoscape "as networks" and have both in exactly the same data structure.

Caveat: Not all graph models are the same! For a topological pathway analysis, you will want to very carefully assess how pathways are modeled by these tools. For example, whether genes and metabolites interact directly or via reaction nodes, which would completely change measures of path length, diffusion, etc. In the case of WikiPathways, there is an RDF model of pathwys, for example, that is different from the Cytoscape model you get via the app. You may want to pull content in via RDF depending on your analysis plan. Or you may want to restructure the imported graphs via the app.

Also check out this paper: https://dx.doi.org/10.3389%2Ffgene.2019.00059 ("Interactions with at least two annotated interaction participants (gene product, metabolite, complex) are included. Gene products have unified Ensembl (Zerbino et al., 2017) identifiers, metabolites have either Wikidata (Mietchen et al., 2015), ChEBI (Hastings et al., 2015) or HMDB identifiers (Wishart et al., 2017), and complexes have Reactome identifiers. ").