Custom Networks
Requirements for Networks in CIE
Networks in CIE must be described by two files. These files are usually written to disk. However, sometimes the network is too large and requires filtering. In this case, the data may be stored in another format and desired portions returned as two data frames by a function. In either case, the two tables must be as follows.
Must consist of the following columns
- uid: an integer which acts as a unique identifier. Synonymous with row number but included as a column for easier operations.
- name: HGNC symbol
- id: Entrez id
- type: biotype, one of Protein or mRNA.
An example of such a table is shown below
 |
Must consist of the following columns
- uid: an integer which acts as a unique identifier. Distinct from the uid in ents. Synonymous with row number but included as a column for easier operations.
- srcuid: the ents uid of the source node, or regulator
- trguid: the ents uid of the target node
- type: type of interaction, must be one of increase, decrease, or conflict. These correspond to up regulation, down regulation, and missing or conflicting information respectively.
An example of such a table is shown below
 |
Additional columns containing metadata about the rels entries may be included. However this is not used by the CIE package in any form of enrichment.
Since custom network file names are not known to the runCIE function, you must supply the rels and ents data frames using the parameters ents and rels.
We used publicly available databases of regulator-target interactions to build our network files. We also used STRINGdb, with the assumption that if a protein targets a protein it also targets that gene. The networks utilized are TRED, TRRUST, STRINGdb, and ChIP Atlas. While we only provide the network files for humans, other species are available through these databases.