• A Package is a collection of functions
• Associated with a given task/analysis/data-type
• The main repository is "The Comprehensive R Archive Network" (https://cran.r-project.org/)

Tools > Install Packages...

• All packages (~1200) are for Bioinformatics
  • Statistical Analysis; Databases & Data Handling; Visualisation
  • NGS data, microarrays, flow cytometry, proteomics…
• New releases every ~6 months
• All packages come with a descriptive vignette

browseVignettes()

Also has an active support community:

• https://support.bioconductor.org/
• Large suite of tutorials & workflows

A recent training course: http://www.bioconductor.org/help/course-materials/2016/BioC2016/

3 Broad Headings based on package tags, or biocViews

1. Software
2. AnnotationData
3. ExperimentData

1. Software

• Currently >1000 packages, primarily for analysis
• Heavily used array packages: affy, gcrma, limma
• Access to external databases: biomaRt, topGO
• Rich in Seq analysis packages: edgeR, DESeq, RSamtools
• Wrappers for external Seq tools: muscle, RBowtie
• Lots of new object classes defined

2. Annotation

• Currently >900 packages
• Set database classes (OrgDb, TxDb, OrganismDb, BSgenome)
• Annotations for common microarrays (e.g. Affy & Illumina)

3. Experiment Data

• Currently ~300 packages
• Includes standard datasets for algorithm testing
• Also those included in many training courses

• Packages don't appear in the drop-down menu for RStudio
  • Tools > Install Packages > ???
• Can be added to your default repositories, but there is a preferred installation procedure

source("http://bioconductor.org/biocLite.R")

• This installs the package BiocInstaller
• Manages the synchronisation of R releases and Bioconductor updates
• The main installation function is biocLite()
• Installs from both CRAN & Bioconductor

R dependencies can be challenging!

To check that you have the tested package versions and fix them

library(BiocInstaller)
biocValid(fix = TRUE)

Let's install some key packages for today

biocLite(c("biomaRt", "AnnotationHub", "GenomicRanges", "rtracklayer"))

• Make up a significant proportion of Bioconductor Packages
• Often seen as the end point of analysis
• For networks/pathways it's the starting point

The package biomaRt is based on the web interface at http://www.ensembl.org/biomart/martview

library(biomaRt)
allMarts <- listMarts()

These are the possible data sources (i.e. marts) we can access

Each mart has multiple datasets

mart <- useMart("ENSEMBL_MART_ENSEMBL")
ensDatasets <- listDatasets(mart)
library(dplyr)
filter(ensDatasets, grepl("sapiens", dataset))

We can just go straight there by selecting the dataset within useMart()

mart <- useMart(biomart = "ENSEMBL_MART_ENSEMBL", 
                dataset = "hsapiens_gene_ensembl")

NB: This is exactly the same procedure as the windows on the web GUI

Now the mart & dataset have been selected

• The main query function is getBM()

?getBM

This will give the requested data directly into a data.frame

The two main pieces of data

• attributes are the values we are looking for
• filter along with values are our search queries

To find what attributes can be downloaded from our mart

martAttributes <- listAttributes(mart)
head(martAttributes, 10)

These are possible pieces of information we can return (dim(martAttributes))

• Some attributes may contain large amounts of data
• We can use filters to restrict our search
• e.g. we may have only a few genes of interest

martFilters <- listFilters(mart)
head(martFilters, 10)

• Let's get all the gene names on Chromosome 1
• NB: We need to specify the filter, and give the filter values separately
• We need to specify the mart argument every time
• Our query field will not be included in the results

genes <- getBM(attributes=c("hgnc_symbol", "entrezgene"), 
               filters = "chromosome_name", 
               values = "1", mart = mart)
head(genes)

ids <- c("ENSG00000134460", "ENSG00000163599")
attr <-  c("ensembl_gene_id", "ensembl_transcript_id")
test <- getBM(attributes = attr,
              filters = "ensembl_gene_id",
              values = ids,
              mart = mart)

Repeat the above without asking for the gene_id back

How could we also get the chromosome, strand, start & end positions in the above query

attr <-  c("ensembl_gene_id", "ensembl_transcript_id", 
           "chromosome_name", "strand", 
           "transcript_start", "transcript_end")
test <- getBM(attributes = attr,
              filters = "ensembl_gene_id",
              values = ids,
              mart = mart)

Here's a problem

?select

We now have more than one function called select

How will R know which one to use"

This is a well known problem

The specific version of a function can be called by using the package name

• Known as the namespace
• dplyr::select() or biomaRt::select()

This session relies heavily on material from

Annotation Resources

Authors: Marc RJ Carlson, Herve Pages, Sonali Arora, Valerie Obenchain, and Martin Morgan

Presented at BioC2015 July 20-22, Seattle, WA

https://github.com/mrjc42/BiocAnnotRes2015

Four common classes of annotation

library(AnnotationHub)
ah <- AnnotationHub()

• This is a relatively new & sensibly named package
• We can access & find numerous annotation types
• Uses SQL-type methods
• Creating this object will create a cache with the latest metadata from each data source

Get a summary:

ah

This is another S4 object

• 3 important components: $dataprovider, $species & $rdataclass
• additional components listed under additional mcols() can also be accessed with the $

We can find the data providers

unique(ah$dataprovider) 

Or the different data classes in the hub

unique(ah$rdataclass) 

We can find the species with annotations

sp <- unique(ah$species)
head(sp)
length(sp)

We can query for matches to any term, e.g. to look for rabbit (Oryctolagus cuniculus) annotation sources

query(ah, "Oryctolagus")

We can create smaller AnnotationHub objects, which we could then search again

We can subset easily

subset(ah, rdataclass=="GRanges")

Or if we know we want the GRanges annotations for the rabbit

subset(query(ah, "Oryctolagus"), rdataclass=="GRanges")

Or we can combine multiple search queries

Fetch the rabbit annotations, which are GRanges objects derived from Ensembl

query(ah, 
      pattern=c("Oryctolagus", "GRanges", "Ensembl"))

We can find the metadata for the whole object, or any subset we've created

meta <- mcols(ah)
meta

There's even a GUI

display(ah)

• You may need to resize the Viewer window
• Return an AnnotationHub() to R by selecting a row & clicking the button

Once we have the specific annotation we're interested in: - subset using the name & the double bracket method - this loads the AnnotationData object into your workspace

gr <- ah[["AH51056"]]
gr

• We'll see more GRanges objects during the week