Pathways, reactions, and biological entities in Reactome knowledge are systematically represented as an ordered network of molecular reactions. Graph database technology is an effective tool for modeling highly connected data, hence Reactome’s relational database is imported in Neo4j to create one large interconnected graph. Instances are represented as nodes and relationships between nodes as edges.
The ReactomeGraph4R
package is an R interface for retrieving data with network structure from the Reactome Graph Database. There is another R package, ReactomeContentService4R
, for querying specific bits of information from the Reactome Database through the RESTful API in the Content Service.
ReactomeGraph4R
is built on the Neo4j driver neo4r
, thus returned data are mainly same as those called by neo4r
but with a little modifications, and are in these two formats:
nodes
and relationships
information that can be used for visualizationThis package will allow you to interact with the data in Reactome’s graph database in R, with the aim of minimizing the number of Neo4j Cypher
queries that the user will need to perform. For example, if you wanted to retrieve any Reactome information associated with the hypothetical identifier ‘123456789’, you can use matchObject(id="123456789")
, which would be equivalent to using the Cypher query MATCH (rgp:ReferenceGeneProduct) WHERE rgp.identifier = "123456789" RETURN rgp
on the Reactome graph database.
Aside from performing basic Cypher queries and formatting the results as R objects, the package also contains functionality that can not be easily performed using Cypher. This includes finding hierarchical data of an instance (for example what Reactions and Pathways a Protein is found in), getting the entire Reaction sequence/context using preceding/following relationships, what role a PhysicalEntity plays in each of its associated Reactions (catalyst, regulator, input, etc.), searching for research papers that are cited in Reactome’s curations, and even displaying network data. Please read on to see detailed instructions for the ReactomeGraph4R
package - it is a flexible package with plenty of useful functionality for the prospective R-Reactome user!
Follow this instruction to download and setup the Reactome Graph Database, then install ReactomeGraph4R
package.
There are two questions needed to be answered for Neo4j server connection when loading the package. You can change the url port if it’s not 7474
. And if the Neo4j authentication is required, the username and password are same as the ones to log in your local Neo4j database.
## Is the url 'http://localhost:7474'? (Yes/no/cancel)
## Does Neo4J require authentication? (yes/No/cancel)
## Successfully connected to the local Reactome Graph Database v76!
The basic function matchObject
allows you to fetch Reactome objects using:
id
: Reactome or non-Reactome identifier (e.g. UniProt id)displayName
: display name of an objectschemaClass
: schema classproperty
: attributes of Reactome objectsrelationship
: relationship between two nodesMoreover, you could specify the argument returnedAttributes
for retrieving only a few attributes of the targeted object; species
for specific species; and limit
for the number of returned objects. Note that this function only returns “row” data.
The “id” input can be either non-Reactome or Reactome identifiers. If you use a non-Reactome id, remember that you must also specify databaseName
since the default one is “Reactome”. For example, to get the Reactome instance associated with a circadian rhythmic gene PER2:
# Retrieve the object of PER2 gene
# NOTE: if you're unsure which database to specify, you have to leave it as NULL
matchObject(id = "PER2", databaseName = NULL)
## $databaseObject
## schemaClass identifier databaseName displayName
## 1 ReferenceDNASequence PER2 COSMIC (genes) COSMIC (genes):PER2 PER2
## dbId geneName url
## 1 11509503 PER2 http://cancer.sanger.ac.uk/cosmic/gene/overview?ln=PER2
Now we know that the database name should be “COSMIC (genes)”! We can also try with a Reactome id “R-HSA-400219”:
## $databaseObject
## schemaClass speciesName oldStId isInDisease releaseDate
## 1 Reaction Homo sapiens REACT_25088 FALSE 2010-09-21
## displayName stIdVersion dbId
## 1 Beta-TrCP1 binds phosphorylated PER proteins R-HSA-400219.1 400219
## name stId category isInferred
## 1 Beta-TrCP1 binds phosphorylated PER proteins R-HSA-400219 binding TRUE
For multiple ids, say you want to get more information for your significantly enriched pathways, you can use function multiObjects
. The speedUp
option can determine to use the doParallel method or not, details see ?multiObjects
.
# retrieve multiple objects
ids <- c('R-HSA-74158', 'R-HSA-1566977', 'R-HSA-3000157', 'R-HSA-3000178', 'R-HSA-216083')
multiObjects(ids)
Instances can also be fetched by their “displayNames”. Do note that spaces and symbols within the name are required. Here we focus on the complex SUMO1:TOP1 in nucleoplasm “SUMO1:TOP1 [nucleoplasm]” in C. elegans:
## $databaseObject
## schemaClass speciesName isInDisease displayName
## 1 Complex Caenorhabditis elegans FALSE SUMO1:TOP1 [nucleoplasm]
## stIdVersion dbId name stId
## 1 R-CEL-4641301.1 10549504 SUMO1:TOP1 R-CEL-4641301
When retrieving instances belonging to one schema class, it’s better specify the argument limit
as well for restricting the number of returned instances. For all available schema classes see Reactome Data Schema. For instance, to get 5 “EntitySets” in human and then return their display names and stId only:
# Get 5 instance in Class EntitySet and return displayName & stId
entity.set <- matchObject(schemaClass = "EntitySet", species = "human",
returnedAttributes = c("displayName", "stId"), limit = 5)
entity.set[["databaseObject"]] # show as dataframe
By specifying the property
, nodes with the given property (or properties), which are actually attributes/slots of Reactome instances, could be returned. Let’s try to get instances that are chimeric and are in disease.
# Get instances with conditions of properties that are stored in a list
matchObject(property = list(isChimeric = TRUE, isInDisease = TRUE), limit = 10)[["databaseObject"]]
The actual Cypher query for this command is MATCH (n1)-[r:relationship]->(n2) RETURN n1,n2
, therefore the n1
and n2
dataframes in the returned list have the same number of rows, and every two rows with the same index are connected with the given relationship.
## $n1
## schemaClass speciesName isInDisease displayName
## 1 Complex Homo sapiens FALSE p-GFAP:EEF1A1 [lysosomal membrane]
## 2 Complex Homo sapiens FALSE p-GFAP:EEF1A1 [lysosomal membrane]
## 3 Complex Rattus norvegicus FALSE p-Gfap:Eef1a1 [lysosomal membrane]
## stIdVersion dbId name isChimeric stId
## 1 R-HSA-9626070.1 9626070 p-GFAP:EEF1A1 FALSE R-HSA-9626070
## 2 R-HSA-9626070.1 9626070 p-GFAP:EEF1A1 FALSE R-HSA-9626070
## 3 R-RNO-9626031.1 9626031 p-Gfap:Eef1a1 FALSE R-RNO-9626031
##
## $n2
## schemaClass speciesName isInDisease
## 1 EntityWithAccessionedSequence Homo sapiens FALSE
## 2 EntityWithAccessionedSequence Homo sapiens FALSE
## 3 EntityWithAccessionedSequence Rattus norvegicus FALSE
## displayName stIdVersion dbId name stId
## 1 p-GFAP [lysosomal membrane] R-HSA-9626054.1 9626054 p-GFAP R-HSA-9626054
## 2 EEF1A1 [lysosomal membrane] R-HSA-9626022.1 9626022 EEF1A1 R-HSA-9626022
## 3 p-Gfap [lysosomal membrane] R-RNO-9626029.1 9626029 p-Gfap R-RNO-9626029
## startCoordinate referenceType endCoordinate
## 1 1 ReferenceGeneProduct 432
## 2 1 ReferenceGeneProduct 462
## 3 1 ReferenceGeneProduct 430
These following functions in the MATCH family provide several commonly used cases that you might be interested in for Reactome data querying.
Reactome data are organized in a hierarchical way: Pathway --> Reaction --> PhysicalEntity
, or sometimes it might be Pathway --> Reaction --> PhysicalEntity --> ReferenceEntity
where the PhysicalEntity has links to external database information via the ReferenceEntity. You could retrieve the hierarchical data of a given Event (Pathway or Reaction) or Entity (PhysicalEntity or ReferenceEntity) using matchHierarchy
. In this example, we’ll take a look at a RNA sequence (PhysicalEntity) “POU5F1 mRNA [cytosol]” with stable identifier “R-HSA-500358”:
# Get hierarchy data of R-HSA-500358
pou5f1.hierarchy <- matchHierarchy(id = "R-HSA-500358", type = "row")
str(pou5f1.hierarchy, max.level = 1)
## List of 4
## $ physicalEntity:'data.frame': 1 obs. of 12 variables:
## $ event :'data.frame': 3 obs. of 13 variables:
## $ upperevent :'data.frame': 2 obs. of 16 variables:
## $ relationships :'data.frame': 7 obs. of 9 variables:
The RNA sequence we specified is in the physicalEntity
dataframe of the result list. It’s directly connected with those Events in the event
dataframe, which are then connected with Events in the upperevent
. Relationships between all these objects are in relationship
dataframe:
This method can find all ReactionLikeEvents (RLEs) connected with a given Pathway by the relationship “hasEvent”. Additionally, the input can be a RLE, the result would be Pathway(s) linked via “hasEvent” together with other RLEs linked with the Pathways(s). Here we focus on a RLE “OAS1 oligomerizes” with identifier “R-HSA-8983688”.
# Find Reactions connected with R-HSA-8983688
rle <- matchReactionsInPathway(event.id = "R-HSA-8983688", type = "row")
## List of 4
## $ reactionLikeEvent :'data.frame': 1 obs. of 12 variables:
## $ pathway :'data.frame': 1 obs. of 14 variables:
## $ otherReactionLikeEvent:'data.frame': 14 obs. of 12 variables:
## $ relationships :'data.frame': 15 obs. of 9 variables:
otherReactionLikeEvent
are RLEs other than “OAS1 oligomerizes” connected with Pathway “OAS antiviral response”.
## [1] "OAS1 binds viral dsRNA"
## [2] "RNASEL cleaves viral ssRNA"
## [3] "OAS2 binds viral dsRNA"
## [4] "RNASEL cleaves cellular ssRNA"
## [5] "OAS2 produces oligoadenylates"
## [6] "ABCE1 binds RNASEL"
## [7] "PDE12 cleaves 2'-5' oligoadenylates "
## [8] "Viral 2',5'-PDE cleaves 2'-5' oligoadenylates "
## [9] "OAS3 binds viral dsRNA"
## [10] "RNASEL binds 2'-5' oligoadenylate"
## [11] "OAS3 produces oligoadenylates"
## [12] "OAS1 produces oligoadenylates"
## [13] "OASL binds DDX58"
## [14] "OAS2 dimerizes"
The contect of these Events can actually be visualized in R using the exportImage
function from the ReactomeContentService4R
package! And it looks the same as that in Pathway Browser. To get the pathway diagram of Pathway “OAS antiviral response” (stId: R-HSA-8983711) that we just retrieved, and highlight the RLE (stId: R-HSA-8983688) that we specified:
## Connecting...welcome to Reactome v82!
With the diagram shown above, we can see that the Reaction highlighted in blue is in the middle of a Reaction cascade, with other RLEs immediately preceding and following it. In order to know what these preceding and following Reactions are, we can use function matchPrecedingAndFollowingEvents
to find RLEs linked via “precedingEvent”. The argument depth
is used to describe the “variable length relationships”, the default value is 1 (i.e. immediately connected); or you can set all.depth = TRUE
for retrieving the whole context. Details see ?matchPrecedingAndFollowingEvents
.
# Retrieve RLE context with depth = 2
rle.context <- matchPrecedingAndFollowingEvents(event.id = "R-HSA-8983688", depth = 2, type = "row")
str(rle.context, max.level = 1)
## List of 4
## $ precedingEvent:'data.frame': 1 obs. of 12 variables:
## $ event :'data.frame': 1 obs. of 12 variables:
## $ followingEvent:'data.frame': 2 obs. of 12 variables:
## $ relationships :'data.frame': 3 obs. of 9 variables:
Usually we query data in a way like parent to child (parent) --> (child)
, where we provide information about the parent. But with the Graph Database, we are able to search in a reverse direction that is child to parent (parent) <-- (child)
with child’s information only. This “child-to-parent” relationship is called Referral. You could carry out the referral fetching by matchReferrals
that supports Classes “Event”, “PhysicalEntity”, “Regulation”, “CatalystActivity”, “ReferenceEntity”, “Interaction”, “AbstractModifiedResidue”. Depth related arguments could also be specified here. More details sees ?matchReferrals
.
We would look at a Regulation “Negative gene expression regulation by ’EGR2 [nucleoplasm]” with dbId “6810147”:
## $Regulation
## schemaClass
## 1 NegativeGeneExpressionRegulation
## displayName stIdVersion
## 1 Negative gene expression regulation by 'EGR2 [nucleoplasm]' R-HSA-6810147.1
## dbId stId
## 1 6810147 R-HSA-6810147
##
## $databaseObject
## schemaClass displayName stIdVersion dbId stId
## 1 BlackBoxEvent HOXB1 gene is transcribed R-HSA-5617454.3 5617454 R-HSA-5617454
## speciesName isInDisease releaseDate name isChimeric
## 1 Homo sapiens FALSE 2015-12-15 HOXB1 gene is transcribed FALSE
## category isInferred
## 1 omitted TRUE
##
## $relationships
## neo4jId type startNode.neo4jId startNode.dbId startNode.schemaClass
## 1 5505941 regulatedBy 1330001 5617454 BlackBoxEvent
## endNode.neo4jId endNode.dbId endNode.schemaClass properties
## 1 1330002 6810147 NegativeGeneExpressionRegulation 1, 4
The dbId of endNode (endNode.dbId
in $relationships
) is exactly the dbId we just specified.
Interactions of a PhysicalEntity (PE) could be retrieved by matchInteractors
. This method begins with finding the ReferenceEntity matched with the PE, then get the Interactions having “interactor” relationship with the ReferenceEntity. For example, to get interactions of “FANCM [nucleoplasm]” with stable id “R-HSA-419535”:
## List of 4
## $ physicalEntity :'data.frame': 1 obs. of 12 variables:
## $ referenceEntity:'data.frame': 1 obs. of 17 variables:
## $ interaction :'data.frame': 7 obs. of 8 variables:
## $ relationships :'data.frame': 8 obs. of 9 variables:
The roles of PhysicalEntities include “input”, “output”, “regulator”, “catalyst”, which are represented as relationships “input” ,“output”, “regulatedBy”, “catalystActivity” respectively. Therefore, we could retrieve instances that are possibly connected with the given PhysicalEntity via these relationships, and see the exact role(s) from the existing relationships. We’ll take a look at a Polymer “HSBP1 oligomer [cytosol]” and input it into matchPEroles
. Either id
or displayName
could be specified.
# Find possible roles of the given PE
roles <- matchPEroles(pe.displayName = "HSBP1 oligomer [cytosol]")
## List of 3
## $ physicalEntity:'data.frame': 6 obs. of 10 variables:
## $ databaseObject:'data.frame': 6 obs. of 12 variables:
## $ relationships :'data.frame': 6 obs. of 9 variables:
## [1] "output"
Diseases related to a PhysicalEntity or an Event could be found using function matchDisease
. In reverse, you can also get PhysicalEntities/Events associated with a Disease.
# Fetch Reactome instances associated with 'neuropathy' in human
matchDiseases(displayName = "neuropathy", species = "human", type = "row")
## $disease
## schemaClass identifier synonym databaseName displayName dbId
## 1 Disease 870 peripheral neuropathy DOID neuropathy 9635395
## name definition
## 1 neuropathy A nervous system disease that is located_in nerves or nerve cells.
## url
## 1 https://www.ebi.ac.uk/ols/ontologies/doid/terms?obo_id=DOID:870
##
## $databaseObject
## schemaClass displayName dbId name
## 1 ChemicalDrug pralidoxime [extracellular region] 9635003 pralidoxime
## isInDisease stIdVersion stId
## 1 TRUE R-ALL-9635003.1 R-ALL-9635003
##
## $relationships
## neo4jId type startNode.neo4jId startNode.dbId startNode.schemaClass
## 1 1031748 disease 263887 9635003 ChemicalDrug
## endNode.neo4jId endNode.dbId endNode.schemaClass properties
## 1 263890 9635395 Disease 1, 0
Given the PubMed id or the title for a paper, Reactome instances related to this paper could be found by matchPaperObjects
. The DatabaseObjects are connected with the LiteratureReference (i.e. paper) via “literatureReference” relationship. Let’s try with a paper “Aggresomes: a cellular response to misfolded proteins”.
# fetch objects by paper title
matchPaperObjects(displayName = "Aggresomes: a cellular response to misfolded proteins", type = "row")
## $literatureReference
## volume schemaClass journal pages year
## 1 143 LiteratureReference J. Cell Biol. 1883-98 1998
## displayName dbId
## 1 Aggresomes: a cellular response to misfolded proteins 9646681
## pubMedIdentifier title
## 1 9864362 Aggresomes: a cellular response to misfolded proteins
##
## $databaseObject
## schemaClass displayName dbId
## 1 Reaction PolyUb-Misfolded proteins bind vimentin to form aggresome 9646679
## speciesName isInDisease releaseDate stIdVersion
## 1 Homo sapiens FALSE 2019-12-10 R-HSA-9646679.2
## name isChimeric
## 1 PolyUb-Misfolded proteins bind vimentin to form aggresome FALSE
## stId category isInferred
## 1 R-HSA-9646679 binding FALSE
##
## $relationships
## neo4jId type startNode.neo4jId startNode.dbId
## 1 46368 literatureReference 12675 9646679
## startNode.schemaClass endNode.neo4jId endNode.dbId endNode.schemaClass
## 1 Reaction 12676 9646681 LiteratureReference
## properties
## 1 1, 0
The ability to view network graphs is definitely a big advantage of a graph database. Fortunately, R has developed into a powerful tool for network analysis. There are a number of R packages targeted network analysis and visualization, therefore we are able to get a graph just like the one in the Neo4j server, and even to set more visualization options!
Don’t forget that results can also be returned in the “graph” format, which are used to create the network visualization in R! This comprehensive tutorial - Network visualization with R (Ognyanova, K., 2019) - walks through each step on the creation of network graphs in R.
Here we will show a couple of examples to generate an interactive network graph after retrieving the specific Reactome graph data. Let’s say we want to visualize the hierarchical data of a ReferenceEntity “UniProt:P33992 MCM5”.
First install and load the following packages.
# install packages
list.pkg <- c("stringr", "visNetwork", "networkD3", "wesanderson")
new.pkg <- list.pkg[!(list.pkg %in% installed.packages()[ ,"Package"])]
if (length(new.pkg)) {
install.packages(new.pkg, repos = "https://cloud.r-project.org/")
}
# load
invisible(suppressPackageStartupMessages(lapply(list.pkg, library, character.only = TRUE)))
We will try the visNetwork
package which visualizes networks using vis.js
javascript library.
# Get graph output data
graph <- matchHierarchy(displayName = "UniProt:P33992 MCM5", databaseName = "UniProt", type = "graph")
relationships <- graph[["relationships"]]
nodes <- graph[["nodes"]]
nodes <- unnestListCol(df = nodes, column = "properties") # unnest the 'properties' column of lists
head(nodes); head(relationships)
# Transform into visNetwork format for nodes & edges
vis.nodes <- data.frame(id = nodes$id,
label = str_trunc(nodes$displayName, 20), # truncate the long names
group = nodes$schemaClass,
title = paste0("<p><b>", nodes$schemaClass, "</b><br>",
"dbId: ", nodes$dbId, "<br>", nodes$displayName, "</p>"))
vis.edges <- data.frame(from = relationships$startNode,
to = relationships$endNode,
label = relationships$type,
font.size = 16,
font.color = 'steelblue')
head(vis.nodes); head(vis.edges)
We are going to change the visual parameters of nodes and edges by adding them as columns in the dataframes. More customizations see the visNetwork
documentation or ?vignette("Introduction-to-visNetwork")
.
# nodes parameters
## get palette colors with package 'wesanderson'
node.colors <- as.character(wes_palette(n = length(unique(vis.nodes$group)), name = "Darjeeling2"))
names(node.colors) <- levels(factor(vis.nodes$group))
## NOTE: don't use `str_replace_all` here since 'TopLevelPathway' & 'Pathway' share the string 'Pathway'
vis.nodes$color.background <- node.colors[as.numeric(factor(vis.nodes$group))] # node color
vis.nodes$color.border <- "lightgray"
## highlight the instance we specified
vis.nodes$color.border[vis.nodes$label == "UniProt:P33992 MCM5"] <- "pink"
vis.nodes$color.highlight.border <- "darkred"
vis.nodes$borderWidth <- 2 # Node border width
# edges parameters
vis.edges$width <- 1.2 # line width
edges.colors <- as.character(wes_palette(n = length(unique(vis.edges$label)), name = "FantasticFox1"))
names(edges.colors) <- unique(vis.edges$label)
vis.edges$color <- str_replace_all(vis.edges$label, edges.colors) # line color
vis.edges$arrows <- "to" # arrows: 'from', 'to', or 'middle'
vis.edges$smooth <- TRUE # should the edges be curved?
# height & width of the plot can be set here
visnet <- visNetwork(vis.nodes, vis.edges, main = "The hierarchy of protein MCM5",
height = "500px", width = "100%")
visnet
Add a drop-down menu:
We can also take a look at another package networkD3
, which generates network graphs using D3 javascript library.
# the node ids MUST be numeric, and start from 0
nodes.idx <- as.character(as.numeric(factor(nodes$id)) - 1)
names(nodes.idx) <- nodes$id
# transform into networkD3 format
d3.edges <- data.frame(source = as.numeric(str_replace_all(relationships$startNode, nodes.idx)),
target = as.numeric(str_replace_all(relationships$endNode, nodes.idx)),
label = relationships$type)
d3.edges <- d3.edges[order(d3.edges$source), ]
d3.nodes <- cbind(idx=as.numeric(nodes.idx), nodes)
d3.nodes <- d3.nodes[order(d3.nodes$idx), ] # the order MUST be consistent with the 'source'
forceNetwork(Links = d3.edges, Nodes = d3.nodes, Source="source", Target="target",
NodeID = "displayName", Group = "schemaClass", Value = "label",
linkColour = "#afafaf", fontSize = 12, zoom = TRUE, legend = TRUE,
Nodesize = 15, opacity = 0.9, charge = -50)
To modify the forceNetwork graph, one can execute custom javascript code with the htmlwidgets
R package, but it won’t be discussed here.
## R version 4.2.1 (2022-06-23)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 20.04.5 LTS
##
## Matrix products: default
## BLAS: /home/biocbuild/bbs-3.16-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.16-bioc/R/lib/libRlapack.so
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=en_GB LC_COLLATE=C
## [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
## [7] LC_PAPER=en_US.UTF-8 LC_NAME=C
## [9] LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] wesanderson_0.3.6 networkD3_0.4
## [3] visNetwork_2.1.2 stringr_1.4.1
## [5] ReactomeContentService4R_1.6.0 ReactomeGraph4R_1.6.0
##
## loaded via a namespace (and not attached):
## [1] tidyselect_1.2.0 xfun_0.34 bslib_0.4.0 purrr_0.3.5
## [5] vctrs_0.5.0 generics_0.1.3 htmltools_0.5.3 yaml_2.3.6
## [9] utf8_1.2.2 rlang_1.0.6 jquerylib_0.1.4 later_1.3.0
## [13] pillar_1.8.1 glue_1.6.2 DBI_1.1.3 foreach_1.5.2
## [17] lifecycle_1.0.3 htmlwidgets_1.5.4 codetools_0.2-18 evaluate_0.17
## [21] knitr_1.40 fastmap_1.1.0 doParallel_1.0.17 httpuv_1.6.6
## [25] curl_4.3.3 parallel_4.2.1 fansi_1.0.3 highr_0.9
## [29] Rcpp_1.0.9 xtable_1.8-4 promises_1.2.0.1 cachem_1.0.6
## [33] magick_2.7.3 jsonlite_1.8.3 mime_0.12 png_0.1-7
## [37] digest_0.6.30 stringi_1.7.8 dplyr_1.0.10 shiny_1.7.3
## [41] getPass_0.2-2 cli_3.4.1 tools_4.2.1 magrittr_2.0.3
## [45] sass_0.4.2 tibble_3.1.8 tidyr_1.2.1 pkgconfig_2.0.3
## [49] ellipsis_0.3.2 data.table_1.14.4 attempt_0.3.1 neo4r_0.1.1
## [53] assertthat_0.2.1 rmarkdown_2.17 httr_1.4.4 rstudioapi_0.14
## [57] iterators_1.0.14 R6_2.5.1 igraph_1.3.5 compiler_4.2.1