---
title: "GWAS catalog: Phenotypes systematized by the experimental factor ontology"
author: "Vincent J. Carey, stvjc at channing.harvard.edu"
date: "August 2015"
output:
  BiocStyle::html_document:
    highlight: pygments
    number_sections: yes
    toc: yes
---

<!--
%\VignetteEngine{knitr::rmarkdown}
%\VignetteIndexEntry{gwascat: exploring GWAS results using the experimental factor ontology}
-->

```{r style, echo = FALSE, results = 'asis'}
BiocStyle::markdown()
suppressPackageStartupMessages({
library(gwascat)
})
```

# Introduction

The EMBL-EBI curation of the GWAS catalog (originated at NHGRI)
includes labelings of GWAS hit records with terms from the EBI Experimental
Factor Ontology (EFO).  The Bioconductor `r Biocpkg("gwascat")`
package includes a `r Biocpkg("graph")` representation of the ontology
and records the EFO assignments of GWAS results in its basic
representations of the catalog.

# Views of the EFO

Term names are regimented.

```{r getg}
library(gwascat)
data(efo.obo.g)
efo.obo.g
sn = head(nodes(efo.obo.g))
sn
```

The nodeData of the graph includes a `def` field.
We will process that to create a data.frame.

```{r procdef}
nd = nodeData(efo.obo.g)
alldef = sapply(nd, function(x) unlist(x[["def"]]))
allnames = sapply(nd, function(x) unlist(x[["name"]]))
alld2 = sapply(alldef, function(x) if(is.null(x)) return(" ") else x[1])
mydf = data.frame(id = names(allnames), concept=as.character(allnames), def=unlist(alld2))
```

We can create an interactive data table for all terms, but for
performance we limit the table size to terms involving the string
'autoimm'.

```{r limtab}
limdf = mydf[ grep("autoimm", mydf$def, ignore.case=TRUE), ]
library(DT)
suppressWarnings({
datatable(limdf, rownames=FALSE, options=list(pageLength=5))
})
```

# Graph operations

The use of the graph representation allows various approaches
to traversal and selection.  Here we examine metadata for a
term of interest, transform to an undirected graph, and obtain
the adjacency list for that term.

```{r lkg}
nodeData(efo.obo.g, "EFO:0000540")
ue = ugraph(efo.obo.g)
neighISD = adj(ue, "EFO:0000540")[[1]]
sapply(nodeData(subGraph(neighISD, efo.obo.g)), "[[", "name")
```

With RBGL we can compute paths to terms from root.
```{r lkggg}
library(RBGL)
p = sp.between( efo.obo.g, "EFO:0000685", "EFO:0000001")
sapply(nodeData(subGraph(p[[1]]$path_detail, efo.obo.g)), "[[", "name")
```

# Connections to the GWAS catalog

The `mcols` element of the `GRanges` instances provided by
gwascat include mapped EFO terms and EFO URIs.

```{r lkef}
data(ebicat38)
names(mcols(ebicat38))
adinds = grep("autoimmu", ebicat38$MAPPED_TRAIT)
adgr = ebicat38[adinds]
adgr
mcols(adgr)[, c("MAPPED_TRAIT", "MAPPED_TRAIT_URI")]
```