With the accumulation of available 'omics' data, developing automatic methods for integrating and analyzing them into a common context has become a key goal in bioinformatics. In this paper, we propose a general framework (called SIPPER) to integrate heterogeneous information from one or several external data sources into a metabolic network, and to perform an automatic, data-independent analysis. Our approach consists in including neighborhood/non-neighborhood information about proteins and genes into a metabolic network, via the admitted paradigm 'gene produces enzyme(s)'. The integration of this neighborhood information uses a distance involving data about genomic location, or gene expression, or connectivity in a PPI network. It may also use any freely chosen notion of distance between genes or proteins. The resulting model is a new (so-called integrated) network. Its low weight paths are the reaction chains of the metabolic network that are catalyzed by groups of closely related enzymes (according to the information the user chooses to integrate). Consequently, integrating heterogeneous knowledge using SIPPER allows us to find associations between genes, enzymes and metabolic reaction chains involved together in the system behavior.