Skip to contents

The OmniPath intercell database annotates individual proteins and complexes, and we combine these annotations with network interactions on the client side, using import_intercell_network. The architecture of this database is complex, aiming to cover a broad range of knowledge on various levels of details and confidence. We can use the intercell_consensus_filter and filter_intercell_network functions for automated, data driven quality filtering, in order to enrich the cell-cell communication network in higher confidence interactions. However, for many users, a simple combination of the most established, expert curated ligand-receptor resources, provided by this function, fits better their purpose.


  curated_resources = c("Guide2Pharma", "HPMR", "ICELLNET", "Kirouac2010", "CellTalkDB",
    "CellChatDB", "connectomeDB2020"),
  cellphonedb = TRUE,
  cellinker = TRUE,
  talklr = TRUE,
  signalink = TRUE,



Character vector of the resource names which are considered to be expert curated. You can include any post-translational network resource here, but if you include non ligand-receptor or non curated resources, the result will not fulfill the original intention of this function.


Logical: include the curated interactions from CellPhoneDB (not the whole CellPhoneDB but a subset of it).


Logical: include the curated interactions from Cellinker (not the whole Cellinker but a subset of it).


Logical: include the curated interactions from talklr (not the whole talklr but a subset of it).


Logical: include the ligand-receptor interactions from SignaLink. These are all expert curated.


Passed to import_post_translational_interactions: further parameters for the interaction data. Should not contain `resources` argument as that would interfere with the downstream calls.


A data frame similar to interactions (network) data frames, the source and target partners being ligand and receptor, respectively.


Some resources are a mixture of curated and bulk imported interactions, and sometimes it's not trivial to separate these, we take care of these here. This function does not use the intercell database of OmniPath, but retrieves and filters a handful of network resources. The returned data frame has the layout of interactions (network) data frames, and the source and target partners implicitly correspond to ligand and receptor. The data frame shows all resources and references for all interactions, but each interaction is supported by at least one ligand-receptor resource which is supposed to based on expert curation in a ligand-receptor context.


lr <- curated_ligand_receptor_interactions()
#> # A tibble: 6,307 × 15
#>    source target source_genesymbol target_genesymbol is_directed is_stimulation
#>    <chr>  <chr>  <chr>             <chr>                   <dbl>          <dbl>
#>  1 Q9GZX6 Q08334 IL22              IL10RB                      1              1
#>  2 P01588 P19235 EPO               EPOR                        1              1
#>  3 Q9H2A7 O00574 CXCL16            CXCR6                       1              1
#>  4 P21583 P10721 KITLG             KIT                         1              1
#>  5 Q07325 P49682 CXCL9             CXCR3                       1              1
#>  6 P13501 P51681 CCL5              CCR5                        1              1
#>  7 P80075 P51681 CCL8              CCR5                        1              1
#>  8 P13236 P51681 CCL4              CCR5                        1              1
#>  9 P05231 P40189 IL6               IL6ST                       1              1
#> 10 P04083 P25090 ANXA1             FPR2                        1              1
#> # ℹ 6,297 more rows
#> # ℹ 9 more variables: is_inhibition <dbl>, consensus_direction <dbl>,
#> #   consensus_stimulation <dbl>, consensus_inhibition <dbl>, sources <chr>,
#> #   references <chr>, curation_effort <dbl>, n_references <dbl>,
#> #   n_resources <int>