In this tutorial we show you how to query interactions from one of the many resources included in OmniPath, we map additional annotations to the data, and combine the two to build a small tissue specific network out of them
We’ll start by importing libraries, first OmnipathR, and dplyr for data wrangling.
library(OmnipathR)
library(dplyr)With the OmnipathR package loaded, we can download a subset of the interaction data. By calling the import_omnipath_annotations function we assign interaction data from the BioGrid resource to the interactions variable restricting it to human interaction data, by selecting the taxonomy ID 9606.
It is important to mention, that the OmnipathR library queries can be replicated in browser too, as they access specific URLs, depending on the parameters we give here.
interactions <- import_all_interactions(
resources = c('SIGNOR'),
organism = 9606
)## Downloaded 11619 interactions.The R query above translates to the following URL for example. Note that only the required resources and taxIDs are selected, but the rest of the query remained intact. Feel free to give it a go in your browser:
The result should look something like below (in R).
interactions %>% tibble()## # A tibble: 11,619 x 17
## source target source_genesymb… target_genesymb… is_directed is_stimulation
## <chr> <chr> <chr> <chr> <int> <int>
## 1 Q13976 Q13507 PRKG1 TRPC3 1 0
## 2 P18031 Q9H1D0 PTPN1 TRPV6 1 0
## 3 P63244 Q9BX84 RACK1 TRPM6 1 0
## 4 Q96QT4 P04083 TRPM7 ANXA1 1 1
## 5 P17612 Q9GZU1 PRKACA MCOLN1 1 0
## 6 Q16539 P49137 MAPK14 MAPKAPK2 1 1
## 7 P49137 O95453 MAPKAPK2 PARN 1 0
## 8 P49757 P46531 NUMB NOTCH1 1 1
## 9 O00548 P46531 DLL1 NOTCH1 1 1
## 10 P31749 O15111 AKT1 CHUK 1 1
## # … with 11,609 more rows, and 11 more variables: is_inhibition <int>,
## # consensus_direction <int>, consensus_stimulation <int>,
## # consensus_inhibition <int>, dip_url <chr>, sources <chr>, references <chr>,
## # curation_effort <int>, dorothea_level <chr>, n_references <int>,
## # n_resources <int>To give these interactions a bit more depth, we can map annotation data to the interactions. To take a look at the available annotation resources in OmniPath, call the get_annotation_resources() function by running the line below, or calling it from the console.
get_annotation_resources()## [1] "Adhesome" "Almen2009" "Baccin2019"
## [4] "CancerGeneCensus" "CancerSEA" "CellCellInteractions"
## [7] "CellPhoneDB" "CellPhoneDB_complex" "ComPPI"
## [10] "CORUM_Funcat" "CORUM_GO" "CSPA"
## [13] "CSPA_celltype" "DGIdb" "DisGeNet"
## [16] "EMBRACE" "Exocarta" "GO_Intercell"
## [19] "GPCRdb" "Guide2Pharma" "HGNC"
## [22] "HPA_secretome" "HPA_subcellular" "HPA_tissue"
## [25] "HPMR" "HPMR_complex" "ICELLNET"
## [28] "ICELLNET_complex" "Integrins" "IntOGen"
## [31] "iTALK" "KEGG-PC" "kinase.com"
## [34] "Kirouac2010" "LOCATE" "LRdb"
## [37] "Matrisome" "MatrixDB" "MCAM"
## [40] "Membranome" "MSigDB" "NetPath"
## [43] "OPM" "Phobius" "Phosphatome"
## [46] "Ramilowski_location" "Ramilowski2015" "SignaLink_function"
## [49] "SignaLink_pathway" "SIGNOR" "Surfaceome"
## [52] "TCDB" "TFcensus" "TopDB"
## [55] "UniProt_family" "UniProt_keyword" "UniProt_location"
## [58] "UniProt_tissue" "UniProt_topology" "Vesiclepedia"
## [61] "Zhong2015"Let’s import tissue enrichment data from the Human Protein Atlas. Calling the import_omnipath_annotations function first we select the proteins we’d like to gather information on, followed by the resources we are pulling the data from.
A toy example below:
We assign the annotations of the proteins TP53 and LMNA from the tissue section of the HPA into the HPA_small variable. The “wide” setting pivots the data from a long format to wide, which gives us a nice table from the queried data.
HPA_small <- import_omnipath_annotations(
proteins = c('TP53', 'LMNA'),
resources = c('HPA_tissue'),
wide = TRUE
)## Downloaded 1549 annotation records.These queries are also URL accessible. This toy query translates to the following: URL: https://omnipathdb.org/annotations?resources=HPA_tissue&proteins=TP53,LMNA&license=academic
Let’s get the unique proteins from SIGNOR into a vector (list)
protein_list <- c(interactions$source, interactions$target) %>% unique()We can pass this into import_omnipath_annotations just like above. To ensure sensible runtimes for this tutorial here we restrict the queried proteins to the first 500 in the list with the [1:500] slice.
HPA_signor <- import_omnipath_annotations(
proteins = protein_list[1:500],
resources = c('HPA_tissue'),
wide = TRUE
) ## Downloaded 325162 annotation records.Now that we have the data, let’s filter it down to a specific case, like thyroid cancer. We filter out rows where the levels of the proteins are not “not detected”, i.e. we only move forward with the ones that are detected.
HPA_thyroid_cancer <- HPA_signor %>%
filter(
tissue == "thyroid cancer",
level != "Not detected"
)We can combine the two datasets (interactions and HPA_thyroid_cancer) with an inner join. This operation returns records that have matching values in both tables. We first map the annotation to the source column (i.e. the initial member of each interaction) and we “pipe” (forward) these results with the %>% symbol into a second join, where we map the results of the first operation to the target column proteins.
thyroid_cancer_network <- inner_join(
interactions, HPA_thyroid_cancer, # here we select the two dataframes
by=c("source"="uniprot")) %>% # the names of the columns to combine them on
inner_join(
HPA_thyroid_cancer, # the first dataframe is the result of the first operation here, second stays the same
by=c("target"="uniprot"), # the names of the columns to combine them on
suffix=c(".source", ".target")) # suffixes, so that we know which column corresponds to which dataset after the join
thyroid_cancer_network %>% tibble()## # A tibble: 966 x 45
## source target source_genesymb… target_genesymb… is_directed is_stimulation
## <chr> <chr> <chr> <chr> <int> <int>
## 1 Q16539 P49137 MAPK14 MAPKAPK2 1 1
## 2 P49757 P46531 NUMB NOTCH1 1 1
## 3 P31749 O15111 AKT1 CHUK 1 1
## 4 O15111 P19838 CHUK NFKB1 1 1
## 5 P38936 P06493 CDKN1A CDK1 1 0
## 6 Q93009 Q00987 USP7 MDM2 1 1
## 7 P60484 Q70Z35 PTEN PREX2 1 0
## 8 Q70Z35 P60484 PREX2 PTEN 1 0
## 9 P30086 P04049 PEBP1 RAF1 1 1
## 10 P45985 P45983 MAP2K4 MAPK8 1 1
## # … with 956 more rows, and 39 more variables: is_inhibition <int>,
## # consensus_direction <int>, consensus_stimulation <int>,
## # consensus_inhibition <int>, dip_url <chr>, sources <chr>, references <chr>,
## # curation_effort <int>, dorothea_level <chr>, n_references <int>,
## # n_resources <int>, genesymbol.source <chr>, entity_type.source <chr>,
## # organ.source <chr>, tissue.source <chr>, level.source <chr>,
## # status.source <chr>, prognostic.source <chr>, favourable.source <chr>,
## # pathology.source <chr>, n_not_detected.source <chr>, n_low.source <chr>,
## # n_medium.source <chr>, n_high.source <chr>, score.source <chr>,
## # genesymbol.target <chr>, entity_type.target <chr>, organ.target <chr>,
## # tissue.target <chr>, level.target <chr>, status.target <chr>,
## # prognostic.target <chr>, favourable.target <chr>, pathology.target <chr>,
## # n_not_detected.target <chr>, n_low.target <chr>, n_medium.target <chr>,
## # n_high.target <chr>, score.target <chr>The result is a thyroid cancer specific interaction network, roughly 2% the size of the original SIGNOR network.
In this tutorial we learned: