Transcript Diapositiva 1

Annotation, Databases,
GO, Pathways,and
all those things
Information on microarray data
• Consists of different
types of informations
– Genes annotations
– Samples annotations
– Genes expression
levels
– Covariates
– Experimental design
– etc
2
Annotation: Relating
probesets to genes
Use of microarray clone
annnotation
• Often, the result of microarray data analysis is a list of
genes.
• The list has to be summarized with respect to its
biological meaning.
• For this, information about the genes and the related
proteins has to be gathered.
– If the list is small (let’s say, 1–30), this is easily done by reading
database information and/or the available literature.
– Sometimes, lists are longer (100s or even 1000s of genes).
Automatic parsing and extracting of information is needed.
4
From clone information to
genes and proteins (1)
• Microarrays are produced using
information on expressed sequences as
EST clones, cDNAs, partial cDNAs etc.
• At the other end, functional information is
generated (and available) for proteins.
• Hence, there is a need to map a clone
sequence ID to a protein ID.
• This is a non-trivial task
5
From clone information to genes
and proteins: a non-trivial task
• First, there are usually hundreds of ESTs (and
several cDNA sequences) that map to the same
gene.
– The Database Unigene tries to resolve this multiplicity
problem by sequence clustering.
– An alternative approach is taken by Locus Link. This
is a quite stable repository of genomic loci, supposed
to be a single gene.
– Since the emphasis is on well-characterised loci,
Locus Link is not complete.
6
From clone information to genes
and proteins: multiple ways to go
• There are other projects like RefSeq (NCBI) or TIGR Gene Indices.
• According to the cross-references available for a certain microarray,
one or the other may be advantageous
7
An example:
The human genome
• With the completion of the human genome sequence,
you’d think that such ambiguities can be resolved. In
fact, that is not the case.
– Part of the problem is due to the fact that it is hard to predict
gene structure (intron/exon) without knowing the entire mRNA
sequence, which happens for about two-thirds of all genes.
– Then, there are errors in the assembly (putting together the
sequence snippets). A typical symptom is that a gene appears to
map to multiple loci on the same chromosome, with very high
sequence similarity.
– But there are also sequences that are nearly indentical, but
duplicated. This has happened not long ago in evolution by
means of transposable elements.
8
The human genome:
Some figures
• Currently, it’s estimated that the human genome contains about
25,000 – 30,000 genes that code for 50,000 – 100,000 different
transcripts (and thus, proteins).
• Unigene (human section) contains 105,680 clusters, but 45,999
• of them are of size 2 or less.
• RefSeq DNA contains 28,097 human sequences.
• ENSEMBL contains 21,787 predicted genes, 31,609 predicted
transcripts.
• Fully computational methods like Genscan produce more than
65,000 predictions.
• Locus Link contains 15,248 genes with known function, and further
6038 genes without function annotation
9
Function annotation
• Probably, the most important thing you want to know is what the
genes or their products are concerned with, i.e. their function.
• Function annotation is difficult:
– Different people use different words for the same function, or
– may mean different things by the same word.
– The context in which a gene was found (e.g. “TGF-induced gene”) may
not be particularly associated with its function.
• Inference of function from sequence alone is error-prone and
sometimes unreliable.
• The best function annotation systems (GO,SwissProt) use human
beings who read the literature before assigning a function to a gene
10
The Gene Ontology
• To overcome some of the problems, an
annotation system has been created: The Gene
Ontology (http://www.geneontology.org).
– It represents a unified, consistent system, i.e. terms
occur only once, and there is a dictionary of allowed
words.
– Furthermore, terms are related to each other: the
hierarchy goes from very general terms to very
detailed ones.
11
Cross-references with GO
• The GO database exists independently
from other annotation databases
• There exist cross-references (GOA)
enabling to relate other annotations with
those contained in GO
12
Bioconductor and annotations
• Annotation information is managed in Bioconductor
through metadata packages
• These packages contain one-to-one and one-to-many
mappings for frequently used chips, especially Affymetrix
• Information available includes gene names, gene
symbol, database accession numbers, Gene Ontology
function description, enzmye classification number (EC),
relations to PubMed abstracts, and others.
• There are several packages implementing functionalities
to deal with annotation information: annotate, Annbuilder,
ontoTools, GOstats and many more
13
Static vs. Dynamic Annotation
Static Annotation:
• Bioconductor packages containing annotation
information that are installed locally on a
computer
• well-defined structure
• reproducible analyses
• no need for network connection
Dynamic Annotation:
• stored in a remote database
• more frequent updates  possibly different
result when repeating analyses
• more information
• one needs to know about the structure of the
database, the API of the webservice etc.
Available Metadata
• EntrezGene
is a catalog of genetic loci that connects curated sequence
information to official nomenclature. It replaced LocusLink.
• UniGene
defines sequence clusters. UniGene focuses on protein-coding
genes of the nuclear genome (excluding rRNA and
mitochondrial sequences).
• RefSeq
is a non-redundant set of transcripts and proteins of known
genes for many species, including human, mouse and rat.
• Enzyme Commission (EC)
numbers are assigned to different enzymes and linked to genes
through EntrezGene.
15
Available Metadata
• Gene Ontology (GO)
is a structured vocabulary of terms describing gene products
according to molecular function, biological process, or cellular
component
• PubMed
is a service of the U.S. National Library of Medicine. PubMed
provides a rich resource of data and tools for papers in journals
related to medicine and health. While large, the data source is not
comprehensive, and not all papers have been abstracted
16
Available Metadata
• OMIM
Online Mendelian Inheritance in Man is a catalog of human genes
and genetic disorders.
• NetAffx
Affymetrix’ NetAffx Analysis Center provides annotation resources
for Affymetrix GeneChip technology.
• KEGG
Kyoto Encyclopedia of Genes and Genomes; a collection of data
resources including a rich collection of pathway data.
• IntAct
Protein Interaction data, mainly derived from experiments.
• Pfam
Pfam is a large collection of multiple sequence alignments and
hidden Markov models covering manycommon protein domains17
and families.
Available Metadata
• Chromosomal Location
Genes are identified with chromosomes, and where appropriate
with strand.
• Data Archives
The NCBI coordinates the Gene Expression Omnibus (GEO);
TIGR provides the Resourcerer database, and the EBI runs
ArrayExpress.
18
Annotation Packages
• An early design decision was to provide metadata on a per chiptype basis (e.g. hgu133a, hgu95av2)
• Each annotation package contains objects that provide mappings
between identifiers (genes, probes, …) and different types of
annotation data
• One can list the content of a package:
> library("hgu133a")
> ls("package:hgu133a")
[1] "hgu133a" "hgu133aACCNUM"
[3] "hgu133aCHR" "hgu133aCHRLENGTHS"
[5] "hgu133aCHRLOC" "hgu133aENTREZID"
[7] "hgu133aENZYME" "hgu133aENZYME2PROBE"
[9] "hgu133aGENENAME" "hgu133aGO"
[11] "hgu133aGO2ALLPROBES" "hgu133aGO2PROBE"
[13] "hgu133aLOCUSID" "hgu133aMAP"
[15] "hgu133aMAPCOUNTS" "hgu133aOMIM"
[17] "hgu133aORGANISM" "hgu133aPATH"
[19] "hgu133aPATH2PROBE" "hgu133aPFAM"
[21] "hgu133aPMID" "hgu133aPMID2PROBE"
[23] "hgu133aPROSITE" "hgu133aQC"
[25] "hgu133aREFSEQ" "hgu133aSUMFUNC_DEPRECATED"
[27] "hgu133aSYMBOL" "hgu133aUNIGENE"
19
A little bit of history...
(the pre-SQL era)
before: hgu95av2
hgu95av2.db
now:
Annotation Packages
• Objects in annotation packages used to be environments,
hash tables for mapping  now things are stored in SQLite
DB
• Mapping only from one identifier to another, hard to reverse
• quite unflexible
• The user interface still supports many of the old
environment-specific interactions:
You can access the data directly using any of the standard
subsetting or extraction tools for environments:
get, mget, $ and [[.
> get("201473_at", hgu133aSYMBOL)
[1] "JUNB"
> mget(c("201473_at","201476_s_at"), hgu133aSYMBOL)
$`201473_at`
[1] "JUNB"
$`201476_s_at`
[1] "RRM1"
> hgu133aSYMBOL$"201473_at"
[1] "JUNB"
> hgu133aSYMBOL[["201473_at"]]
[1] "JUNB"
21
Working with Metadata
Suppose we are interested in the gene BAD.
> gsyms <- unlist(as.list(hgu133aSYMBOL))
> whBAD <- grep("^BAD$", gsyms)
> gsyms[whBAD]
1861_at 209364_at
"BAD" "BAD"
> hgu133aGENENAME$"1861_at"
[1] "BCL2-antagonist of cell death"
22
Working with Metadata
Find the pathways that BAD is associated with.
> BADpath <- hgu133aPATH$"1861_at"
> kegg <- mget(BADpath, KEGGPATHID2NAME)
> unlist(kegg)
01510
"Neurodegenerative Disorders"
04012
"ErbB signaling pathway"
04210
"Apoptosis"
04370
…
"Colorectal cancer"
05212
"Pancreatic cancer"
05213
"Endometrial cancer"
05215
23
Working with Metadata
We can get the GeneChip probes and the unique EntrezGene loci
in each of these pathways. First, we obtain the Affymetrix IDs
> allProbes <- mget(BADpath, hgu133aPATH2PROBE)
> length(allProbes)
[1] 15
> allProbes[[1]][1:10]
[1] "206679_at" "209462_at" "203381_s_at"
"203382_s_at"
[5] "212874_at" "212883_at" "212884_x_at"
"200602_at"
[9] "211277_x_at" "214953_s_at"
> sapply(allProbes, length)
01510 04012 04210 04370 04510 04910 05030 05210
05212 05213
85 169 162 137 413 243 39 167 156 111
05215 05218 05220 05221 05223
24
194 137 160 117 110
Working with Metadata
And then we can map these to their Entrez Gene values.
> getEG = function(x) unique(unlist(mget(x,
hgu133aENTREZID)))
> allEG = sapply(allProbes, getEG)
> sapply(allEG, length)
01510 04012 04210 04370 04510 04910 05030
05210 05212 05213
37 84 81 67 187 130 18 82 72 51
05215 05218 05220 05221 05223
85 68 74 53 53
25
.db Packages
• Data in the new .db annotation packages is stored in
SQLite databases
 much more efficient and flexible
• old environment-style access provided by objects of class
Bimap (package AnnotationDbi)
left
object
right
object
left
object
right
object
left
object
right
object
26
.db Packages
• Data in the new .db annotation packages is stored in
SQLite databases
 much more efficient and flexible
• old environment-style access provided by objects of class
Bimap (package AnnotationDbi)
left
object
left
object
right
object
name
left
object
attr1 = value1
attr2 0 value2
right
object
right
object
 bipartite graph
27
DBI
• collection of classes and methods for database interaction
• they abstract the particular implementations of common
standard operations on different types of databases
• resultSet: operations are performed on the database, the user
controls how much information is returned
dbSendQuery
create result set
dbGetQuery
get all results
dbGetQuery(connection, sql query)
28
.db Packages
Notice that there are a few more entries here. They give you
access to a connection to the database.
> library("hgu133a.db")
> ls("package:hgu133a.db")
[1] "hgu133aACCNUM" "hgu133aALIAS2PROBE"
[3] "hgu133aCHR" "hgu133aCHRLENGTHS"
[5] "hgu133aCHRLOC" "hgu133aENTREZID"
[7] "hgu133aENZYME" "hgu133aENZYME2PROBE"
[9] "hgu133aGENENAME" "hgu133aGO"
[11] "hgu133aGO2ALLPROBES" "hgu133aGO2PROBE"
[13] "hgu133aMAP" "hgu133aMAPCOUNTS"
[15] "hgu133aOMIM" "hgu133aORGANISM"
[17] "hgu133aPATH" "hgu133aPATH2PROBE"
[19] "hgu133aPFAM" "hgu133aPMID"
[21] "hgu133aPMID2PROBE" "hgu133aPROSITE"
[23] "hgu133aREFSEQ" "hgu133aSYMBOL"
[25] "hgu133aUNIGENE" "hgu133a_dbInfo"
[27] "hgu133a_dbconn" "hgu133a_dbfile"
[29] "hgu133a_dbschema"
29
> con <- hgu133a_dbconn()
> q1 <- "select symbol from gene_info“
> head(dbGetQuery(con ,q1))
symbol
1
A2M
2
NAT1
3
NAT2
4 SERPINA3
extract information from a database table as data.frame
> toTable(hgu133aSYMBOL)[1:3,]
probe_id symbol
1 217757_at
A2M
2 214440_at
NAT1
3 206797_at
NAT2
reverse mapping
> revmap(hgu133aSYMBOL)$BAD
[1] "1861_at"
"209364_at"
30
Lkeys, Rkeys: Get left and right keys of a Bimap object
> head(Lkeys(hgu133aSYMBOL))
[1] "1007_s_at" "1053_at"
"117_at"
"121_at"
"1255_g_at" "1294_at"
> head(Rkeys(hgu133aSYMBOL))
[1] "A2M"
"NAT1"
"NAT2"
"SERPINA3" "AADAC"
"AAMP"
nhit: number of hits for every left key in a Bimap object
> table(nhit(revmap(hgu133aSYMBOL)))
1
2
3
4
5
6
7
8
9
10
11
12
13
18
19
8101 2814 1273 475 205
77
3
4
1
2
1
1
19
15
5
31