An R package for comparing and clustering PRIDE Archive
projects.
This package makes use of the prideR package.
First, we need to install devtools:
install.packages("devtools")
library(devtools)
Then we just call
install_github(username="Bioanalytics", repo="prideRcompare")
The following code
archive.accession.distance.retriever("PXD001034", project.count=100)
Returns the Jaccard similarity index
between the PRIDE Archive project with accession PXD001034 and the first 100
projects returned by the PRIDE Archive web service (that currently are the 100
most recent ones).
We can use ProjectDetail objects in a similar way.
Individual distances can be calculated using distance.ProteinDetail like in
p1034.protein.details <- get.list.ProteinDetail("PXD001034")
p1156.protein.details <- get.list.ProteinDetail("PXD001156")
distance.ProteinDetail(p1034.protein.details, p1156.protein.details)
Also applicable to lists using the function distance.list.ProteinDetail and
distance.df.ProteinDetail.
We can cluster lists of ProteinDetail and ProjectSummary instances, although
in the end the clustering is always done on the protein details and therefore
the later method uses the former one through the PRIDE Archive web service. In
order to be a good citizen, the recommended usage is through ProteinDetail
obtaining first the list of lists of protein details for each project.
cancer.projects.100 <- search.list.ProjectSummary("cancer", 100)
cancer.projects.100.protein.details.100 <- lapply(cancer.projects.100, function(x) {get.list.ProteinDetail(accession(x), 100)})
cancer.clusters.100.100 <- cluster.ProteinDetails(cancer.projects.100.protein.details.100)
That will give us a hierarchical cluster objects (as generated by hclust) that
we can use to find out clusters (e.g. 5 clusters) using:
cancer.projects.100.accessions <- sapply(cancer.projects.100, accession)
cutree.labels(cancer.clusters.100.100, 5, cancer.projects.100.accessions)
Or just plot:
plot(cancer.clusters.100.100, cancer.projects.100.accessions, main="Clustering of latest 100 cancer projects")
Some clustering examples can be found here.