InterMineR Vignette"

In InterMineR: R Interface with InterMine-Powered Databases

knitr::opts_chunk$set(echo = TRUE)

Introduction

InterMine is a powerful open source data warehouse system integrating diverse biological data sets (e.g. genomic, expression and protein data) for various organisms. Integrating data makes it possible to run sophisticated data mining queries that span domains of biological knowledge. A selected list of databases powered by InterMine is shown in Table 1:

Database | Organism | Data | ---------|----------|------| FlyMine | Drosophila | Genes, homology, proteins, interactions, gene ontology, expression, regulation, phenotypes, pathways, diseases, resources, publications HumanMine | H. sapiens | Genomics, SNPs, GWAS, proteins, gene ontology, pathways, gene expression, interactions, publications, disease, orthologues, alleles MouseMine | M. musculus | Genomics, proteins, gene ontology, expression, interactions, pathways, phenotypes, diseases, homology, publications RatMine | R. norvegicus | Disease, gene ontology, genomics, interactions, phenotype, pathway, proteins, publication QTL, SNP WormMine | C. elegans | Genes, alleles, homology, go annotation, phenotypes, strains YeastMine | S. cerevisiae | Genomics, proteins, gene ontology, comparative genomics, phenotypes, interactions, literature, pathways, gene expression ZebrafishMine | D. rerio | Genes, constructs, disease, gene ontology, genotypes, homology, morpholinos, phenotypes TargetMine | H. sapiens, M. musculus | Genes, protein structures, chemical compounds, protein domains, gene function, pathways, interactions, disease, drug targets MitoMiner | H. sapiens, M. musculus, R. norvegicus, D. rerio, S. cerevisiae, S. pombe | Genes, homology, localisation evidence, Mitochondrial reference gene lists, phenotypes, diseases, expression, interactions, pathways, exome IndigoMine | Archae | Genomics ThaleMine | A. thaliana | Genomics, proteins, domains, homology, gene ontology, interactions, expression, publications, pathways, GeneRIF, stocks, phenotypes, alleles, insertions, TAIR MedicMine | Medicago truncatula | Genomics, pathways, gene ontology, publications, proteins, homology PhytoMine | over 50 plant genomes | Genes, proteins, expression, transcripts, homology

Please see the InterMine home page for a full list of available InterMines.

InterMine includes an attractive, user-friendly web interface that works 'out of the box' and a powerful, scriptable web-service API to allow programmatic access to your data. This R package provides an interface with the InterMine-powered databases through Web services.

Jumpstart: How to build queries using InterMineR

Let's start with a simple task - find the pathways of the gene ABO.

Select a database

First, we look at what databases are available.

library(InterMineR)
listMines()

Since we would like to query human genes, we select HumanMine.

# load HumaMine
im <- initInterMine(mine=listMines()["HumanMine"])
im

Obtain a prebuilt query

Both in InterMine database website and in InterMineR, you are able to build custom queries. However, to facilitate the retrieval of information from InterMine databases, a variety of pre-built queries, called templates, have also been made available. Templates are queries that have already been created with a fixed set of output columns and one or more constraints.

# Get template (collection of pre-defined queries)
template = getTemplates(im)
head(template)

We would like to find templates involving genes.

# Get gene-related templates
template[grep("gene", template$name, ignore.case=TRUE),]

The template Gene_Pathway seems to be what we want. Let's look at this template in more detail.

# Query for gene pathways
queryGenePath = getTemplateQuery(
  im = im, 
  name = "Gene_Pathway"
)
queryGenePath

There are three essential members in a query - SELECT, WHERE and constraintLogic.

1. SELECT a. The SELECT (or view) represents the output columns in the query output. b. Columns of a view are usually of the form "A.B", where B is the child of A. For example in the column Gene.symbol, symbol is the child of Gene. Columns could also be in cascade form "A.B.C". For example, in the column Gene.locations.start, locations is the child of Gene and start is the child of locations.
2. WHERE a. The WHERE statement is a collection of constraints. b. Query constraints include a list of the following columns: i. path 1. in the same format as view columns ii. op 1. the constraint operator 2. Valid values: "=", "!=", "LOOKUP", "ONE OF", "NONE OF", ">", "<", ">=", "<=", "LIKE" iii. value 1. the constraint value iv. code 1. Ignore 2. The logic code for the constraint (e.g. A, B or C). 3. Only used in the constrainLogic (discussed below v. extraValue 1. optional, required for LOOKUP constraints 2. Short name of organism, e.g. H. sapiens vi. Editable 1. Ignore 2. Used to determine if user is allowed to edit this constraint. Only for the UI. vii. Switchable 1. Ignore 2. Used to determine if user is allowed to disable this constraint.
   Only for the UI. viii. Switched 1. Ignore 2. Used to determine if user has enabled this constraint. Only for the UI.
3. constraintLogic a. Constraint Logic, if not explicitly given, is "AND" operation, e.g., "A and B", where A and B are the codes in the constraints.

Look at the data model

What does 'Gene.symbol' mean? What is 'Gene.pathway.identifier'?

Let's take a look at the data model. NOTE: Section temporarily removed due to errors

```r

model <- getModel(im)

head(model)

Let's look at the children of the Gene data type.

# ```r

# model[which(model$type=="Gene"),]

Gene has a field called 'symbol' (hence the column Gene.symbol). Gene also references the Pathways class, which is of the Pathway data type.

```r

model[which(model$type=="Pathway"),]

## Run a Query
Let's now run our template.

```r

resGenePath <- runQuery(im, queryGenePath)
head(resGenePath)

Modify a Query

Edit a constraint

Let's modify the query to find the pathways of the gene ABO. We want to change the 'value' attribute from PPARG to ABO.

There are two ways to build a query in InterMineR.

1. We can either build a query as a list object with newQuery function, and assign all input values (selection of retrieved data type, constraints, etc.) as items of that list,

2. Or we can build the query as an InterMineR-class object with the functions setConstraint, which allows us to generate a new or modify an existing list of constraints, and setQuery, which generates the query as a InterMineR-class object.

setConstraints and setQuery functions are designed to facilitate the generation of queries for InterMine instances and avoid using multiple iterative loops, especially when it is required to include multiple constraints or constraint values (e.g. genes, organisms) in your query.

# modify directly the value of the first constraint from the list query
queryGenePath$where[[1]][["value"]] <- "ABO"

# or modify the value of the first constraint from the list query with setConstraints
queryGenePath$where = setConstraints(
  modifyQueryConstraints = queryGenePath,
  m.index = 1,
  values = list("ABO")
)

queryGenePath$where

Note the value is now equal to 'ABO'. Let's rerun our query with the new constraint.

resGenePath <- runQuery(im, queryGenePath)
head(resGenePath)

Now we are seeing pathways for the ABO gene.

Add a new constraint

You can also add additional filters. Let's look for a specifc pathway.

There are four parts of a constraint to add:

1. path a. I got the path from the output columns but I could have figured out it from the data model.
2. op a. Valid values: "=", "!=", "LOOKUP", "ONE OF", "NONE OF", ">", "<", ">=", "<=", "LIKE"
3. value a. What value I am filtering on.
4. code a. Must be a letter not in use by the query already. Looking at the query output above we can see we only have one constraint, labelled 'A'. Let's use 'B' for our code.

newConstraint <- list(
  path=c("Gene.pathways.name"),
  op=c("="), 
  value=c("ABO blood group biosynthesis"), 
  code=c("B")
)

queryGenePath$where[[2]] <- newConstraint
queryGenePath$where

Our new filter has been added successfully. Rerun the query and you see you only have one pathway,ABO blood group biosynthesis, returned.

resGenePath <- runQuery(im, queryGenePath)
resGenePath

Add a column

You can also add additional columns to the output. For instance, is the Gene also involved in any disease? Let's add this information.

Let's see what we know about diseases.

```r

model[which(model$type=="Gene"),]

```

The Gene data type has an 'Diseases' reference of type 'Disease'.

```r

model[which(model$type=="Disease"),]

```

Disease has an attribute called "name". Add Gene.diseases.name to the view. We'll add it as the last column, we can see from above there are 7 other columns already so we'll put it as #8:

# use setQuery function which will create an InterMineR-class query
queryGenePath.InterMineR = setQuery(
  inheritQuery = queryGenePath,
  select = c(queryGenePath$select, 
             "Gene.diseases.name")
  )

getSelect(queryGenePath.InterMineR)
#queryGenePath.InterMineR@select

# or assign new column directly to the existing list query
queryGenePath$select[[8]] <- "Gene.diseases.name"
queryGenePath$select

# run queries
resGenePath.InterMineR <- runQuery(im, queryGenePath.InterMineR)
resGenePath <- runQuery(im, queryGenePath)

all(resGenePath == resGenePath.InterMineR)

head(resGenePath, 3)

NB: adding columns can result in changing the row count.

Change constraint logic

The constraintLogic, if not given, is 'A and B'. We would now try to explicitly specify the constraintLogic. A and B corresponds to the 'code' for each constraint.

queryGenePath$constraintLogic <- "A and B"
queryGenePath$constraintLogic

Run the query again and see no change:

resGenePath <- runQuery(im, queryGenePath)
resGenePath

Change to be 'A or B' and see how the results change.

Recipes

Obtain the gene ontology (GO) terms associated with gene ABO

• Start with the template Gene GO

queryGeneGO <- getTemplateQuery(im, "Gene_GO")
queryGeneGO

• Modify the view to display a compact view

queryGeneGO$select <- queryGeneGO$select[2:5]
queryGeneGO$select

• Modify the constraints to look for gene ABO.

queryGeneGO$where[[1]][["value"]] <- "ABO"
queryGeneGO$where

• Run the query

resGeneGO <- runQuery(im, queryGeneGO )
head(resGeneGO)

Obtain the genes associated with gene ontology (GO) term 'metal ion binding'

• Start with the template Gene GO

queryGOGene <- getTemplateQuery(im, "GOterm_Gene")
queryGOGene

• Modify the view to display a compact view

queryGOGene$select <- queryGOGene$select[2:5]
queryGOGene$select

• Modify the constraints to look for GO term 'metal ion binding'

queryGOGene$where[[1]]$value = "metal ion binding"
queryGOGene$where

• Run the query

resGOGene <- runQuery(im, queryGOGene )
head(resGOGene)

Find and plot the genes within 50000 base pairs of gene ABCA6

• Start with the Gene_Location template, update to search for ABCA6 gene.

queryGeneLoc = getTemplateQuery(im, "Gene_Location")
queryGeneLoc$where[[2]][["value"]] = "ABCA6"
resGeneLoc= runQuery(im, queryGeneLoc)

resGeneLoc

We're going to use the output (gene location) as input for the next query.

• Define a new query

# set constraints
constraints = setConstraints(
  paths = c(
    "Gene.chromosome.primaryIdentifier",
    "Gene.locations.start",
    "Gene.locations.end",
    "Gene.organism.name"
  ),
  operators = c(
    "=",
    ">=",
    "<=",
    "="
  ),
  values = list(
    resGeneLoc[1, "Gene.chromosome.primaryIdentifier"],
    as.character(as.numeric(resGeneLoc[1, "Gene.locations.start"])-50000),
    as.character(as.numeric(resGeneLoc[1, "Gene.locations.end"])+50000),
    "Homo sapiens"
  )
)

# set InterMineR-class query
queryNeighborGene = setQuery(
  select = c("Gene.primaryIdentifier", 
             "Gene.symbol",
             "Gene.chromosome.primaryIdentifier",
             "Gene.locations.start", 
             "Gene.locations.end", 
             "Gene.locations.strand"),
  where = constraints
)

summary(queryNeighborGene)

• Run the query

resNeighborGene <- runQuery(im, queryNeighborGene)
resNeighborGene

• Plot the genes

resNeighborGene$Gene.locations.strand[which(resNeighborGene$Gene.locations.strand==1)]="+"

resNeighborGene$Gene.locations.strand[which(resNeighborGene$Gene.locations.strand==-1)]="-"

gene.idx = which(nchar(resNeighborGene$Gene.symbol)==0)

resNeighborGene$Gene.symbol[gene.idx]=resNeighborGene$Gene.primaryIdentifier[gene.idx]

require(Gviz)

annTrack = AnnotationTrack(
  start=resNeighborGene$Gene.locations.start,
  end=resNeighborGene$Gene.locations.end,
  strand=resNeighborGene$Gene.locations.strand,
  chromosome=resNeighborGene$Gene.chromosome.primaryIdentifier[1],
  genome="GRCh38", 
  name="around ABCA6",
  id=resNeighborGene$Gene.symbol)

gtr <- GenomeAxisTrack()
itr <- IdeogramTrack(genome="hg38", chromosome="chr17")

plotTracks(list(gtr, itr, annTrack), shape="box", showFeatureId=TRUE, fontcolor="black")

System info

sessionInfo()

Appendix

Visual way to derive the column name of a query view or the path name in a query constraint from the database webpage

The InterMine model could be accessed from the mine homepage by clicking the tab "QueryBuilder" and selecting the appropriate data type under "Select a Data Type to Begin a Query":

Here we select Gene as the data type:

We could select Symbol and Chromosome->Primary Identifier by clicking Show on the right of them. Then click "Export XML" at the bottom right corner of the webpage:

The column names Gene.symbol and Gene.chromosome.primaryIdentifier are contained in the XML output:

InterMineR documentation built on Nov. 8, 2020, 5:58 p.m.