[BioC] graph::join discards node and edge attributes

[Nishant Gopalakrishnan]

Hi Paul,

The new graph classes are already available in the graph package for
testing.  There is also a  vignette MultiGraphClass.Rnw that runs
through some examples in using the graphBAM and MultiGraph classes to
get you started.

Nishant

On 12/01/2010 01:55 PM, Paul Shannon wrote:
> Hi Nishant,
>
> This is very good news -- I like your solution.
>
> Could you venture out on a limb and predict when the new graph classes will be ready for testing?  They have features I'd like to use in the next (or some subsequent) version of RCytoscape.
>
>  - Paul
>
>
> On Dec 1, 2010, at 1:48 PM, Nishant Gopalakrishnan wrote:
>
>   
>> Hi Paul,
>>
>> The join/union/ intersect methods for the graph objects do not preserve
>> the edge or the node attributes. I do not think there are any plans to
>> add this feature the old graph classes.
>>
>> However, the new graphBAM class has methods that will hopefully do what
>> you are looking for . The graphUnion method performs union of two graphs
>> while preserving the node and edge attributes.  If there is a conflict
>> of attributes, it is filled with NA. You do have the option of resolving
>> the conflict by providing a named list of functions to the edgeFun or
>> nodeFun arguments of the graphUnion method.
>> The function should
>>  1. take in two arguments and returns a conflict resolved value for the
>> attribute.
>>  2. have the same name as the attribute
>>
>> I shall illustrate with some simple examples for the edge attributes below
>>
>> ### Create two  graphs g1 and g2  with color and weight edge attributes
>> from = c("a", "b", "d", "d")
>> to   = c("b", "c", "y", "x")
>> weight=c(1.2, 2.4, 5.4, 3.2)
>> df <- data.frame(from, to, weight)
>> g1 <- graphBAM(df, edgemode = "directed")
>> edgeData(g1, attr = "color") <- "green"  ## set the default
>> edgeData(g1, from = c("a","b"), to = c("b", "c") , attr = "color") <-
>> c("yellow", "blue")
>>
>> from = c("a", "b", "b", "d", "d")
>> to   = c("b", "c", "d", "c", "x")
>> weight=c(1.2, 4.2, 5.6, 2.1, 3.2)
>> df <- data.frame(from, to, weight)
>> g2 <- graphBAM(df, nodes = c("a","b","c", "d", "x", "y", "z"),
>>    edgemode = "directed")
>> edgeData(g2, attr = "color") <- "green"
>> edgeData(g2, from =  c("a","b"), to = c("b", "c") , attr = "color") <-
>> c("yellow", "cyan")
>>
>> ## confilicting attributes are filled with NA in the result
>> g <- graphUnion(g1, g2)  
>> edgeData(g, attr = "weight")  
>> edgeData(g, attr = "color")
>>
>> ## Provide a function for resolving conflicting weight attribute, I want
>> the sum of
>> ## weights in case of conflict
>> g <- graphUnion(g1, g2, edgeFun = list(weight = sum))
>> edgeData(g, attr = "weight")
>> edgeData(g, attr = "color")
>>
>> If you wanted to resolve the conflict for the attribute color for the
>> edges, you could write a function
>> myFun <- function(x, y) {
>>  ## resolve conflict here
>> }
>> and pass it  an argument to the graphUnion method.
>> g <- graphUnion(g1, g2, edgeFun = list(weight = sum, color = myFun))
>>
>> Node attributes if present are also handled in exactly the same manner.
>> If conflicting node attributes occur in the union process , there is
>> another argument  nodeFun that you can use to pass in a function for
>> resolving the conflict.
>>
>> I hope that answers your question.
>>
>> Nishant
>>
>> On 11/29/2010 09:11 AM, Paul Shannon wrote:
>>     
>>> I have two well-annotated kegg pathway graphs I wish to merge.
>>>
>>> Using graph::join, the nodes and edges merge nicely, but all the edge and node attributes are lost, and a new edge attribute 'weight' is added.
>>>
>>>
>>>       
>>>> names (nodeDataDefaults (g.05200))  
>>>>
>>>>         
>>>   [1] "type"  "label" "desc"  "BP"    "MF"    "CC"    "phase" "score" "lfc"   "pval"
>>>
>>>       
>>>> names (nodeDataDefaults (g.04010))  
>>>>
>>>>         
>>>   [1] "type"  "label" "desc"  "BP"    "MF"    "CC"    "phase" "score" "lfc"   "pval"
>>>
>>>       
>>>> names (nodeDataDefaults (g.both))   
>>>>
>>>>         
>>>  NULL
>>>
>>>
>>>
>>>       
>>>> names (edgeDataDefaults (g.04010))
>>>>
>>>>         
>>>  [1] "edgeType"   "pmid"       "abstract"   "reciprocal" "score"      "source"
>>>
>>>       
>>>> names (edgeDataDefaults (g.05200))
>>>>
>>>>         
>>>  [1] "edgeType"   "pmid"       "abstract"   "reciprocal" "score"      "source"
>>>
>>>       
>>>> names (edgeDataDefaults (g.both))
>>>>
>>>>         
>>>  [1] "weight"
>>>
>>> Is there any graph method, current or planned, which will preserve node and edge attributes?  I realize that attribute conflicts can arise, and some strategy is required to handle these.
>>>
>>> - Paul
>>>
>>>
>>>
>>>       
>>>> sessionInfo ()
>>>>
>>>>         
>>> R version 2.12.0 (2010-10-15)
>>> Platform: x86_64-apple-darwin9.8.0/x86_64 (64-bit)
>>>
>>> locale:
>>> [1] en_US.utf-8/en_US.utf-8/C/C/en_US.utf-8/en_US.utf-8
>>>
>>> attached base packages:
>>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>>
>>> other attached packages:
>>> [1] RCytoscape_1.1.6     org.Mm.eg.db_2.4.1   RCurl_1.4-3          bitops_1.0-4.1       GOstats_2.16.0       Category_2.16.0      GO.db_2.4.5
>>> [8] GSEABase_1.11.2      annotate_1.28.0      KEGG.db_2.4.1        KEGGgraph_1.4.0      graph_1.26.0         XML_3.2-0            SPIA_1.8.0
>>> [15] org.Hs.eg.db_2.4.6   RSQLite_0.9-2        DBI_0.2-5            AnnotationDbi_1.11.8 Biobase_2.10.0       RUnit_0.4.26
>>>
>>> loaded via a namespace (and not attached):
>>> [1] genefilter_1.32.0 RBGL_1.25.0       splines_2.12.0    survival_2.35-8   tools_2.12.0      XMLRPC_0.2-2      xtable_1.5-6
>>> _______________________________________________
>>> Bioconductor mailing list
>>> Bioconductor at r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>>> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
>>>
>>>       
>>     
>