In svenbioinf/OGRE: Calculate, visualize and analyse overlap between genomic regions

knitr::include_graphics('./logo.png')

knitr::opts_chunk$set(tidy = FALSE,
                      cache = FALSE,
                      dev = "png",
                      message = FALSE, error = FALSE, warning = TRUE)
vigDir=getwd()
if(dir.exists("../inst/extdata")){
  knitr::opts_knit$set(root.dir = "../inst/extdata")
}else{
  knitr::opts_knit$set(root.dir = "../extdata")
}

Installation

Install OGRE using Bioconductor's package installer.

if(!requireNamespace("BiocManager", quietly = TRUE))
  install.packages("BiocManager")
BiocManager::install("OGRE")

Load the OGRE package:

library(OGRE)

Quick start- load datasets from hard drive

To start up OGRE you have to generate an OGREDataSet that is used to store your datasets and additional information about the analysis that you are conducting. Query and subjects files can be conveniently stored in their own folders as GenomicRanges objects in form of stored .rds / .RDS files. We point OGRE to the correct location by supplying a path for each folder with the character vectors queryFolder and subjectFolder. In this vignette we are using lightweight query and subject example data sets to show OGRE's functionality.

myQueryFolder <- file.path(system.file('extdata', package = 'OGRE'),"query")
mySubjectFolder <- file.path(system.file('extdata', package = 'OGRE'),"subject")

myOGRE <- OGREDataSetFromDir(queryFolder=myQueryFolder,
                             subjectFolder=mySubjectFolder)

By monitoring OGRE's metadata information you can make sure the input paths you supplied are stored correctly.

metadata(myOGRE)

Query and subject datasets are read by loadAnnotations() and stored in the OGREDataSet as GRanges objects. We are going to read in the following example datasets:

• query "genes" (242 Protein coding genes)
• subject "CGI" (365 CpG islands)
• subject "TFBS" (48761 Transcription factor binding sites)

myOGRE <- loadAnnotations(myOGRE)

OGRE uses your dataset file names to label query and subjects internally, we can check these names by using the names() function since every OGREDataSet is a GRangesList.

names(myOGRE)

Let's have a look at the stored datasets:

myOGRE

To find overlaps between your query and subject datasets we call fOverlaps(). Internally OGRE makes use of the GenomicRanges package to calculate full and partial overlap as schematically shown.

knitr::include_graphics(file.path(vigDir,'overlap.png'))

Any existing subject - query hits are then listed in detailDT and stored as a data.table.

myOGRE <- fOverlaps(myOGRE)
head(metadata(myOGRE)$detailDT,n=2)

The summary plot provides us with useful information about the number of overlaps between your datasets.

 myOGRE <- sumPlot(myOGRE)
 metadata(myOGRE)$barplot_summary

Using the Gviz visualization each query can be displayed with all overlapping subject elements. Choose labels for all region tracks by supplying a trackRegionLabels vector. Plots are stored in the same location as your dataset files.

 myOGRE <- gvizPlot(myOGRE,"ENSG00000142168",showPlot = TRUE,
                    trackRegionLabels = setNames(c("name","name"),c("genes","CGI")))

The overlap distribution can be generated with summarizeOverlap(myOGRE) and outputs a table with informative statistics such as minimum, lower quantile, mean, median, upper quantile, and maximum number of overlaps per region and per dataset. Overlap distribution can also be displayed as histograms using plotHist(myOGRE) and accessed by metadata(myOGRE)$hist and metadata(myOGRE)$summaryDT. Two tables / plots are generated. The first one showing numbers for regions with and without overlap and the second one showing numbers only for regions with overlap by excluding all others. Next, we generate an histogram with the number of TFBS per gene (x-axis, log scale) and the TFBS frequency (y-axis). When focusing only on regions with overlap, we see that genes have on average (median) 54 TFBS overlaps (black dashed line).

 myOGRE <- summarizeOverlap(myOGRE) 
 myOGRE <- plotHist(myOGRE)
 metadata(myOGRE)$summaryDT
 metadata(myOGRE)$hist$TFBS

It is possible to create an average coverage profile of all gene-TFBS overlaps, split in 100 bins, which represent gene bodies of all 242 genes. Both, forward and reverse coding genes are arranged on the x-Axis and peaks indicate an TFBS overlap enrichment. Overlap coverage is calculated as the sum of all gene TFBS overlaps in 5'-3'direction. Generated plots can be accessed by metadata(myOGRE)$covPlot$TFBS and the resulting profile shows an accumulation of TFBS around gene start and end positions.

 myOGRE <- covPlot(myOGRE) 
 metadata(myOGRE)$covPlot$TFBS$plot

Quick start- load datasets from AnnotationHub

AnnotationHub offers a wide range of annotated datasets which can be manually aquired but need some parsing to work with OGRE as detailed in vignette section Frequently Asked Questions(FAQ). For convenience addDataSetFromHub() adds one of the predefined human datasets of listPredefinedDataSets() to an OGREDataSet. Those are taken from AnnotationHub and are ready to use for OGRE. We start by creating an empty OGREDataSet and attaching one dataset after another, whereby one query and two subjects are added. The datasets are now ready for further analysis.

myOGRE <- OGREDataSet() 
listPredefinedDataSets()
myOGRE <- addDataSetFromHub(myOGRE,"protCodingGenes","query")
myOGRE <- addDataSetFromHub(myOGRE,"CGI","subject")
myOGRE <- addDataSetFromHub(myOGRE,"TFBS","subject")
names(myOGRE)

As you can see, the three datasets proteinCodingGenes, CGI and TFBS are stored within OGRE. You can then continue with overlap analysis using fOverlaps().

Quick start- load user defined GenomicRanges (GRanges) datasets

To offer more flexibility addGRanges() enables the user to attach additional datasets to OGRE in form of GenomicRanges objects. Again we start by creating an empty OGREDataSet and generate an example GenomicRanges object which is then added to your OGREDataSet either as "query" or "subject".

myOGRE <- OGREDataSet()
myGRanges <- makeExampleGRanges()
myOGRE <- addGRanges(myOGRE,myGRanges,"query")

Frequently asked questions

How to add additional datasets from AnnotationHub?

Use AnnotationHub() to connect to AnnotationHub. Each dataset is stored under a unique ID and can be accessed in a list like fashion i.e. aH[["AH5086"]]. Queries like c("GRanges","Homo sapiens", "CpG") enable browsing through datasets. In this case we are searching for human CpG islands ranges stored as GenomicRanges objects. For more information refer to ?AnnotationHub To make those datasets compatible with OGRE additional parsing is needed as stated in [How to add custom GenomicRanges datasets?]

aH <- AnnotationHub()
aH[["AH5086"]]
q <- query(aH, c("GRanges","Homo sapiens", "CpG"))

How to add custom GenomicRanges datasets?

Any GenomicRanges datasets can be added that fulfill basic compatibility requirements. GenomicRanges objects must:

• Originate from a common genome build i.e. "HG19"

Use GenomeInfoDb::genome() on any GenomicRanges object to get/set genome information

• Contain the same set of chromosomes i.e. chr1 != 1 or chrM != MT

Use GenomeInfoDb::seqinfo() on any GenomicRanges object to get/set chromosome information

• Contain a "name" and a (unique) "ID" column

Use S4Vectors::mcols() on any GenomicRanges object to get/set metadata information

How to add datasets stored as .gff files?

Datasets from external sources are often stored as .gff (v2&v3) files. Once those files exist in the query or subject folder and their attribute columns contain "ID" and "name" information, OGRE tries to load them. Working example .gff files can be found on OGRE's github page in folder: inst/extdata/gffTest.

myOGRE <- OGREDataSetFromDir(queryFolder = "pathToQueryFolder",
                             subjectFolder = "pathToSubjectFolder")
myOGRE <- loadAnnotations(myOGRE)

How to add datasets stored as tabular files?

Datasets stored as tabular files like .csv or .bed may need some preprocessing for them work with OGRE. We recommend reading them in with read.table() or data.table::fread() to obtain a data frame object. After making sure the dataset complies with the requirements in section [How to add custom GenomicRanges datasets?], GenomicRanges::makeGRangesFromDataFrame() offers a convenient way to generate GenomicRanges object from data frames.

What type of overlaps are reported?

Both, partial overlap, where only a part of two (or more) regions are overlapping and complete overlap, where one region is completely overlapped by another, are reported.

How to change dataset names?

OGRE automatically infers dataset names based on your file names. You can either change your file names before you start OGRE or you can use names(myOGRE) <- c("NewName1", "NewName2","...") after you read in your datasets.

Session info

sessionInfo()

svenbioinf/OGRE documentation built on Jan. 30, 2023, 10:55 p.m.