The publicly available interaction databases have non-standard protein identifications, file formats and are not uniquely indexed and annotated, which compromises the development of a single algorithm to integrate all datasets. The interacting pairs constructed by the method described above may be error prone and must undergo a validation step. In order to achieve a more reliable result, some facts should be considered: proteins that actually interact are expected to share the same cellular compartment and have common interaction partners. It has been shown that a pair of genuine interacting proteins is generally expected to have a common cellular role and proteins that have common interaction partners have a higher chance of sharing a common function. Moreover, even if two proteins are Cynarin consistently predicted to interact they must be located at the same cell compartment and at the same time. The interactions in the GPMGDID present a Class score similar to the cellular compartment classification described by Branda˜o et al., and it is based on three characteristics: type of interaction, number of papers describing the interaction in PubMed, and cellular component described for the Isochlorogenic-acid-C interacting nodes in the Gene Ontology database. The FSWeight approach was initially designed to predict protein functions, and lately has shown a good performance in evaluating the reliability of protein interactions. The interaction pairs of proteins that are classified with high score by this method are likely to be true positives. On the other hand, the pairs of proteins that are classified with low scores are likely to be false positives. The most interesting feature of the FSWeight is that it is able to rank the reliability of an interaction between a pair of proteins using only the topology of the interactions between that pair of proteins and their neighbors within a short radius in a graph network. Therefore, we implemented in GPMGDID the Functional Similarity Weight score calculation originally proposed by Chua et al., and described by Branda˜o et al., for all first, second and third level interactions present in our database. The effect of FSW score threshold in the network is exemplified and discussed in the Results and Discussion section.