User:Nbertin/Transcript model derived annotation protocol
Proposed here is an annotation of promoter based on their overlap with model transcripts (RefSeq for all organisms and gencode for human samples) feature : proximal promoter region, 5' and 3' UTR, exon (coding, located in UTR region or associated to non-coding transcripts), intron (similarly coding, located in UTR region or associated to non-coding transcripts).
The protocol and script described here use a combination of calls to BedTools (intersectBed, groupBy, etc) to annotate clusters produced by F5 HelicosCAGE clustering pipeline and formatted as an OSCfile (see User:Tlassmann/clustering for the details of the clustering pipeline) with respect to an ordered serie of BED6 formatted annotations. This script will be run as part of the CAGE post-processing pipeline and data file for each independant libraries will be provided in datasets updates.
Alternatively you can also use the script CAGE-Tag-Cluster-Annotation.sh to redo this annotation step with your own clustered data or using a different list of annotations, altering their ordering, etc or to obtain promoter / 'gene' expression aggregated over the provided annotations.
Promoter annotations of "UPDATE_012 Decomposition-based Peak Identification (DPI) cluster "
Annotation datasets
At the exception of F5_human_lncRNAome and TBP_JASPAR_CORE_MA0108.2, all source files where obtained from
- F5's UCSC mirror :
- F5's ZENBU instance
RefSeq, Ensembl and gencode
Note that Gencode corresponds to gencode V7 (added by Kawaji-san in GencodeV7, while 2011-06-20 UCSC dumpstill contains gencodeV4)
Split non-coding and protein-coding separately according to the following sub-categories :
INPUT.non_coding.first_exon.bed INPUT.non_coding.inner_exon.bed INPUT.non_coding.last_exon.bed
INPUT.non_coding.first_intron.bed INPUT.non_coding.inner_intron.bed INPUT.non_coding.last_intron.bed
INPUT.non_coding.tss.bed INPUT.non_coding.tss_upstream1000.bed INPUT.non_coding.tss_upstream100.bed
INPUT.non_coding.tss_upstream500.bed INPUT.protein_coding.tss.bed INPUT.protein_coding.first_exon.bed
INPUT.protein_coding.inner_exon.bed INPUT.protein_coding.last_exon.bed INPUT.protein_coding.first_exon_in_3UTR.bed
INPUT.protein_coding.inner_exon_in_3UTR.bed INPUT.protein_coding.last_exon_in_3UTR.bed INPUT.protein_coding.first_exon_in_5UTR.bed
INPUT.protein_coding.inner_exon_in_5UTR.bed INPUT.protein_coding.last_exon_in_5UTR.bed INPUT.protein_coding.first_exon_in_CDS.bed
INPUT.protein_coding.inner_exon_in_CDS.bed INPUT.protein_coding.last_exon_in_CDS.bed INPUT.protein_coding.first_intron.bed
INPUT.protein_coding.inner_intron.bed INPUT.protein_coding.last_intron.bed INPUT.protein_coding.first_intron_in_3UTR.bed
INPUT.protein_coding.inner_intron_in_3UTR.bed INPUT.protein_coding.last_intron_in_3UTR.bed INPUT.protein_coding.first_intron_in_5UTR.bed
INPUT.protein_coding.inner_intron_in_5UTR.bed INPUT.protein_coding.last_intron_in_5UTR.bed INPUT.protein_coding.first_intron_in_CDS.bed
INPUT.protein_coding.inner_intron_in_CDS.bed INPUT.protein_coding.last_intron_in_CDS.bed INPUT.protein_coding.tss_upstream1000.bed
INPUT.protein_coding.tss_upstream100.bed INPUT.protein_coding.tss_upstream500.bed
In addition to non-coding and protein-coding, gencode annotations also contain the pseudo-gene class
lncRNA (provided by Leonard)
F5_human_lncRNAome(Jia&Lipovich_Gencode)BED.zip
INPUT.first_exon.bed INPUT.inner_exon.bed INPUT.last_exon.bed INPUT.first_intron.bed INPUT.inner_intron.bed INPUT.last_intron.bed INPUT.tss.bed INPUT.tss_upstream1000.bed INPUT.tss_upstream100.bed INPUT.tss_upstream500.bed
UCSC ESTs
Overlapping ESTs exonic sequences have been merged, only the name of upt to 3 ESTs are reported
INPUT.exon.bed INPUT.intron.bed INPUT.tss.bed
INPUT.tss_upstream1000.bed INPUT.tss_upstream100.bed INPUT.tss_upstream500.bed
UCSC all_mrna
INPUT.first_exon.bed INPUT.inner_exon.bed INPUT.last_exon.bed INPUT.first_intron.bed INPUT.inner_intron.bed INPUT.last_intron.bed INPUT.tss.bed INPUT.tss_upstream1000.bed INPUT.tss_upstream100.bed INPUT.tss_upstream500.bed
CpG island
CpG island coordinates extracted from F5 UCSC mirror were expanded 200bp downstream in order to replicate the analysis in F3 Nat. Gen. :
"... we determined whether a ... CpG (within 200 bp) upstream of the start site of the clusters was present..."
TATA box
TATA box coordinate were extracted from Michiel's JASPAR_CORE TBP_MA0108.2 Position Weight Matrix whole genome scanning
Only JASPAR_CORE TBP_MA0108.2 PWM with a ppvalue greater than 3.5 were retained.
This threshold might need to be revisited after Sebastian and Timo's analysis of TSS surrounding motifs. An overview of the impact of this threshold on gencode coding and non-codoing derived TSS can be seen in CAGE-Tag-Cluster-Annotation-Building/TBP_JASPAR_CORE_MA0108.2.hg19.scanning_vs_gencodev7.tss.ppval_distrib.pdf
UCSC Repeat masker
repName, repClass and repFamily annotate each cluster
Annotation Set Building pipeline
Details of how (and scripts) the annotations sets have been built can be found in
- CAGE-Tag-Cluster-Annotation-Building.readme
- CAGE-Tag-Cluster-Annotation-Building
- In particular the script CAGE-Tag-Cluster-Annotation-Building.split_bed12.pl is useful to break down a transcript BED12 files into sub components (intron|exon, first|inner|last 5UTR|CDS|3UTR)
Annotation Pipeline
Human Decomposition-based Peak Identification (DPI) cluster
Mouse Decomposition-based Peak Identification (DPI) cluster
Annotation Results
Human Decomposition-based Peak Identification (DPI) cluster
- https://fantom5-collaboration.gsc.riken.jp/webdav/home/nbertin/CAGE-Tag-Cluster-Annotation_Aug11/tc.decompose_smoothing_merged.hg19.annotations/
- tc.decompose_smoothing_merged.hg19.CpGislands.annotated.osc.gz
- tc.decompose_smoothing_merged.hg19.EST.annotated.osc.gz
- tc.decompose_smoothing_merged.hg19.Ensembl.non_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.hg19.Ensembl.protein_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.hg19.F5_human_lncRNAome.annotated.osc.gz
- tc.decompose_smoothing_merged.hg19.RefSeq.non_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.hg19.RefSeq.protein_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.hg19.TBP_JASPAR_CORE_MA0108.2.annotated.osc.gz
- tc.decompose_smoothing_merged.hg19.gencode-pseudo.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.hg19.gencode.non_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.hg19.gencode.protein_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.hg19.knownGene.non_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.hg19.knownGene.protein_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.hg19.mRNA.annotated.osc.gz
- tc.decompose_smoothing_merged.hg19.rmsk.annotated.repClass.repFamily.osc.gz
Mouse Decomposition-based Peak Identification (DPI) cluster
- https://fantom5-collaboration.gsc.riken.jp/webdav/home/nbertin/CAGE-Tag-Cluster-Annotation_Aug11/tc.decompose_smoothing_merged.mm9.annotations/
- tc.decompose_smoothing_merged.mm9.CpGislands.annotated.osc.gz
- tc.decompose_smoothing_merged.mm9.EST.annotated.osc.gz
- tc.decompose_smoothing_merged.mm9.Ensembl.non_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.mm9.Ensembl.protein_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.mm9.RefSeq.non_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.mm9.RefSeq.protein_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.mm9.TBP_JASPAR_CORE_MA0108.2.annotated.osc.gz
- tc.decompose_smoothing_merged.mm9.knownGene.non_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.mm9.knownGene.protein_coding.annotated.sym.osc.gz
- tc.decompose_smoothing_merged.mm9.mRNA.annotated.osc.gz
- tc.decompose_smoothing_merged.mm9.rmsk.annotated.repClass.repFamily.osc.gz
Annotating your own clusters
Using CAGE-Tag-Cluster-Annotation.sh
File:CAGE-Tag-Cluster-Annotation.sh.zip
CAGE-Tag-Cluster-Annotation.sh
File:CAGE-Tag-Cluster-Annotation-Bin.zip
CAGE-Tag-Cluster-Annotation.sh accompanying required BedTools binaries
CAGE-Tag-Cluster-Annotation.sh annotation files
CAGE-Tag-Cluster-Annotation.sh annotation files
Please note that this is now version 1.8
which has significant speed when improvement dealing with already BED-like formatted input and also allows for numbering the column according to their input order (annotated output will have a different order corresponding to those of the ordered matching annotations)
usage: ./CAGE-Tag-Cluster-Annotation.sh -o <OSCfile to annotate> -a <file listing the annotation to be used>
Use a combination of calls to BedTools (intersectBed, groupBy, etc) to annotate clusters
produced by F5 HelicosCAGE clustering pipeline and formatted as an OSCfile (-o or STDIN for
F5 HelicosCAGE clustering pipeline output files which assumes a precise ordering of genome
coodinates columns (see details of the -o option below) and the addition of tab-delmimted
feature-wise metadata, or -b for BED like ordering of genome coodinates columns, note that
it will also accomodate additioan tab delimited columns see details of the -b option below)
with respect to an ordered list of BED6 formatted annotations files, those files are to
be listed in a tab delimited text file (-a) which along with the path to the actual BED6
formatted annotation file contains information regarding the respective orientation of
the cluster and annotation to be considered (aka in 'sense' or 'anitsense'), the name to
be reported and if clusters overlapping a first annotation should be filtered out or taken
into consideration or not if overlapping a second annotation further down the ordered list
(for more details, see the description the -a option or few of the illustrative examples
below).
Additionally, it is possible to trigger the summation of (numerical) columns (-s) by
aggregating over the annotation names (-n) and/or annotation classes (-c) of the list of
BED6 formatted annotations files (See examples below for more details).
version: 1.8
OPTIONS:
-h Show this message.
-v Verbose level (1-9).
-w Show the commented lines of this script.
SIMPLE CLUSTER ANNOTATIONS
-o Path to the (optionally gzip-compressed) OSCfile to be annotated
The OSCfile is likely obtained from F5 HelicosCAGE clustering pipeline,
therefore, it is assumed that :
* The file contains a header, each line of which is marked by "##".
* Comment line that will be reported but whose position relative to
the data might be altered are marked by "#".
* The first non "#|##" line contains a short description of the column
content.
* Columns are tab delimited.
* The ordering of the columns is fixed and corresponds to :
id, chrom, start.0base, end, strand, raw.<exp_name>, norm.<exp_name>
See the -b parameter for alternative BED compliant column ordering
Note: actually works as long as ordering of the 1st columns are
<something>, chrom, start.0base, end, strand
* Positions are 0-based (see column ordering expectations).
* 'chrom' names must be similar to those defined in the annotation files.
* The file may be in a gzip-compressed format.
-b Path to the (optionally gzip-compressed) BED6-like formatted OSCfile.
The -b option is an alternative to using F5 HelicosCAGE clustering pipeline OSCfile,
in which the ordering of the columns is fixed and corresponds to :
chrom, start.0base, end, id, score, strand
Note that the column order is preseved in the output annotation added OSCfile
Similarly to the -o parameter: 'chrom' names must be similar to those of the annotation
files, the BED6-like file can contain comment lines marked by '#' and can be in a
gzip-compressed form.
-a Path to the file containing the list of sorted BED6 annotation file paths
This list is to be sorted in the order in which potential annotations need to be
looked-up, this ordering is crucial and allows for prioritizing one of several
possible annotations, for example a cluster matching the promoter of a transcript
may also overlap the intron of another transcript,this alternative annotation being
somehow less relevant than the first one (note that this can be turned of by setting
the filter flag as anything but '1', see below).
* This file must be a tab delimited txt file with the following 4 columns
1) A name for the annotation (aka prom500,S gencode_exon,AS ...) which
describes synthetically the content of this annotation.
This string will be appended to the annotated OSCFile send to STDOUT
in its one-before-last column.
2) The path to the BED6 formatted file holding the annotation
3) The relative orientation of the annotation and input data (sense or
antisense).
Accepted values are "sense", "1", "S", or "antisense", "-1", "AS".
4) A flag set to "1" for the filtering out of the overlapping data in
subsequent sequential annotation look up.
* This file may contain comments marked by the character "#" that will be ignored
when parsed for execution of this script
* The basename of this file will be used as the header associated metadata in :
- ParameterValue[annotation_list_path] = <basename of this file>
- ParameterValue[annotation_list] = <basename of this file>
- ColumnVariable[annotation.<basename of this file>.class] = the class of ...
- ColumnVariable[annotation.<basename of this file>.names] = comma delimited ...
-g Flag allowing the grouping of annotation names sharing the same class (aka prom, UTR)
It is recommended to use this flag in order to obtain a single concise annotation for
each given cluster.
-z Flag triggering the reporting of the sole annotation classes (aka without annotation names)
-x allows for specifying the string used to mark "intergenic" aka cluster that did not
overlap with any of the listed annotation.
-y Adds a column corresponding the (comment and header less) line number of the input OSC or
BED6-like file. This option is useful to recover the initial sorting of the input file from
the annotated outputed OSCfile (which otherwise is sorted in part along the order of the
matched annotations provided by the -a parameter)
This line number is added just after the last column of the input data and right before
the annotations
the ColumnVariable[InputFile.DataLineNum] = <line_number> is added to the output OSCheader
ADDITIONAL OPTIONS FOR CLUSTER/ANNOTATIONS-DERIVED EXPRESSION LEVEL
-s Comma delimited list of the column index to be summed up. In which case, the flag -n, -c
or -nc becomes mandatory
-n Flag triggering the aggregation using the overlapping annotation name (aka NM_2345)
-c Flag triggering the aggregation using the class of the overlapping annotation (aka
prom500,S UTR,AS etc). Note that giving the list of comma delimited column indexes to
be summed up (option -s) becomes mandatory.
BUG NEEDING FIXING
[Feb 2nd 2011] When piping successive calls to ./CAGE-Tag-Cluster-Annotation.sh,
blank lines are inserted between the header and the actual data.
while this is not an issue for most usage of the script, it may cause
problem when parsing the file into ZENBU
EXAMPLES
1) ./CAGE-Tag-Cluster-Annotation.sh -o cluster.osc -a refgene.some_annotion_list.txt -g > cluster.refgene_annotated.osc
Will Annotate clusters with respect to refgene coding and non-coding, proximal
promoter regions, introns, exons, UTRs, ...
Clusters (XYZ and ABC) overlapping the proximal promoter regions of more than
one transcripts (say NM_12345 and NM_23456 or NM_56789) will be annotated as
"XYZ XYZ_coord prom500,S NM_12345,NM_23456"
"ABC ABC_coord exon,S NM_56789,"
2) ./CAGE-Tag-Cluster-Annotation.sh -o cluster.osc -a refgene.some_annotation_list.txt > cluster.refgene_split_annotation.osc
Same as example 1) but ommiting the -g flag
clustres will be annotated as "XYZ XYZ_coord prom500,S NM_12345"
"XYZ XYZ_coord prom500,S NM_23456"
"ABC ABC_coord exon,S NM_56789,"
3) ./CAGE-Tag-Cluster-Annotation.sh -o cluster.osc -a refgene.some_annotion_list.txt -g | ./CAGE-Tag-Cluster-Annotation.sh \
-a repeat.annotation_list.txt -g -x NA> cluster.refgene_and_repeat_annotated.osc
By piping the result of example 1) into a second round of annotations
(e.g. repeat element), with "NA" (specified using -x) when no overlap
was found clusters will be annotated as
"XYZ XYZ_coord prom500,S NM_12345,NM_23456 repeat LINE"
"XYZ XYZ_coord prom500,S NM_23456 repeat LINE"
"ABC ABC_coord exon,S NM_56789, NA NA"
4) ./CAGE-Tag-Cluster-Annotation.sh -o cluster.osc -a refgene.prom_UTR.txt -n -s 6,7 > cluster.refgene_expression.osc
Provided refgene_prom_UTR.txt lists BED a file of refgene promoters
and a BED file of UTRs. This will allow the gathering of the gene
expression level in tagcount and tmp (6th and 7th column in test.osc),
is other word summing up all the expression of annotation overlapping
clusters. Resulting in an OSC file containing "NM_12345 50 0.25"
"NM_23456 26 0.13"
5) ./CAGE-Tag-Cluster-Annotation.sh -o cluster.osc -a refgene.prom_UTR.txt -nc -s 6,7 > cluster.refgene_subfeature_expression.osc
Same as above but adding the flag -c
expression will be reported as "prom500,S NM_12345 45 0.225"
"5UTR,S NM_12345 4 0.02"
"5UTR,AS NM_12345 1 0.005"
"prom500,S NM_23456 25 0.0125"
"5UTR,S NM_12345 1 0.005"
Using the latest BedTools "annotateBed" tool
The latest version of Bedtools (v2.11.2, 31st January) includes a novel tool called "annotateBed" that annotates one BED/VCF/GFF file with the coverage and number of overlaps observed from multiple other BED/VCF/GFF files. In this way, it allows one to ask to what degree one feature coincides with multiple other feature types with a single command. This is a potentially good alternative to CAGE-Tag-Cluster-Annotation.sh provided the file you want to annotate is BED VCF or GFF formatted and you only care about one quantitative value (that you would bury in the BED6::score columns).
Here is a quick way to transform the OSCfile provided in FANTOM5 updates into BED and a (rather simplistic) example of annotateBed usage, counting the number of entries of each annotation for each cluster (note that here the score is et to "1" but could be replaced by any of the column containing a quantitative value of interest, also for more clever use of the tool please have a look at BedTools man page)
awk '{FS="\t"}{OFS="\t"}{print $2,$3,$4,$1,"1",$5}' cluter_file.osc > cluster_file.bed
annotateBed -s -counts -i cluster_file.bed -files refseq_prom500.bed refseq_exon.bed refseq_intron.bed > cluster_file.annotated
Side note, since the ordering of the data in cluster_file.bed should be preserved in cluster_file.annotated, to get antisense count, one may use :
awk '{FS="\t"}{OFS="\t"}{print $2,$3,$4,$1,"1",$5}' file.osc > file.bed
annotateBed -counts -i file.bed -files refseq_prom500.bed refseq_exon.bed refseq_intron.bed > file.annotated_SAS
annotateBed -s -counts -i file.bed -files refseq_prom500.bed refseq_exon.bed refseq_intron.bed > file.annotated_S
paste file.annotated_SAS file.annotated_S | awk '{FS="\t"}{OFS="\t"}{print $1,$2,$3,$4,$5,$6,$7-$14,$8-$15}' > file.annotated_AS
Promoter / 'gene' expression
Complementing the annotation of promoters based on their overlap with model transcripts (RefSeq, Gencode) feature (proximal promoter region, UTR, exon, intron), an extension of the script which aggregates the value of a particular OSCfile given the associated annotation category can be used to obtain promoter or 'transcript' expression levels from an input OSCfile
This extension allows for "complex" transcript expression level to be obtained
i.e. report the exprxpression level of a transcript as the sum up the tags/cluster overlapping the proximal promoter and 5'UTR exons only.
Or as the sum up the tags/cluster overlapping the proximal promoter and entire first 5'UTR exon and intron, etc
usage: ./CAGE-Tag-Cluster-Annotation.sh -o <OSCfile to annotate> -a <file listing the annotation to be used>
Use a combination of calls to BedTools (intersectBed, groupBy, etc) to annotate clusters
produced by F5 HelicosCAGE clustering pipeline and formatted as an OSCfile (-o or STDIN for
F5 HelicosCAGE clustering pipeline output files...
Additionally, it is possible to trigger the summation of (numerical) columns (-s) by
aggregating over the annotation names (-n) and/or annotation classes (-c) of the list of
BED6 formatted annotations files (See examples below for more details).'''
version: 1.8
OPTIONS:
-h Show this message.
-v Verbose level (1-9).
-w Show the commented lines of this script.
SIMPLE CLUSTER ANNOTATIONS
-o Path to the (optionally gzip-compressed) OSCfile to be annotated
-b Path to the (optionally gzip-compressed) BED6-like formatted OSCfile.
-a Path to the file containing the list of sorted BED6 annotation file paths
-g Flag allowing the grouping of annotation names sharing the same class (aka prom, UTR)
It is recommended to use this flag in order to obtain a single concise annotation for
each given cluster.
ADDITIONAL OPTIONS FOR CLUSTER/ANNOTATIONS-DERIVED EXPRESSION LEVEL
-s Comma delimited list of the column index to be summed up. In which case, the flag -n, -c
or -nc becomes mandatory
-n Flag triggering the aggregation using the overlapping annotation name (aka NM_2345)
-c Flag triggering the aggregation using the class of the overlapping annotation (aka
prom500,S UTR,AS etc). Note that giving the list of comma delimited column indexes to
be summed up (option -s) becomes mandatory.
EXAMPLES
1) ./CAGE-Tag-Cluster-Annotation.sh -o cluster.osc -a refgene.prom_UTR.txt -n -s 6,7 > cluster.refgene_expression.osc
Provided refgene_prom_UTR.txt lists BED a file of refgene promoters
and a BED file of UTRs. This will allow the gathering of the gene
expression level in tagcount and tmp (6th and 7th column in test.osc),
is other word summing up all the expression of annotation overlapping
clusters. Resulting in an OSC file containing "NM_12345 50 0.25"
"NM_23456 26 0.13"
2) ./CAGE-Tag-Cluster-Annotation.sh -o cluster.osc -a refgene.prom_UTR.txt -nc -s 6,7 > cluster.refgene_subfeature_expression.osc
Same as above but adding the flag -c
expression will be reported as "prom500,S NM_12345 45 0.225"
"5UTR,S NM_12345 4 0.02"
"5UTR,AS NM_12345 1 0.005"
"prom500,S NM_23456 25 0.0125"
"5UTR,S NM_12345 1 0.005"
Future directions / Suggestions
Add repeat elements (UCSC rmsk table) to the list of provided default annotationsAdd CpG island (from UCSC cpgIslandExt table)Add ESTs (obtained from UCSC all_est table)Add TATA box (in particular take advantage of Michiel Motif Activity derived TPB scanning to define TATA box)- for TATA box and CpG island, write a script / wrapper around BedTools "closestBed" to not only provide the CpG/TATA status of promoters but more informatively give the distance to which the closest CpG/TATA is located
- few ENCODE supertracks
- Julian Gough's group protein domains annotations (via Ensmbl transcripts)
Do not hesitate to add your own suggestions here (this is a wiki page) or email me at nbertin@gsc.riken.jp
- ...
- ...
(DEPRECATED) promoter annotations "up to UPDATE_011"
Annotation datasets
RefSeq
Concerns were raised over the completness of RefSeq annotation, In future UPDATES, it will be replaced by ensembl
Ensembl
Instead of RefSeq for cross-species annotation / expression comparison
Gencode (hg19 only)
lncRNA (provided by Leonard)
The current reference nonredundant list of 8858 human lncRNA genes (chr, direction, TSS, end): Media:F5_human_lncRNAome(Jia&Lipovich_Gencode_Lander).xls
FANTOM5 annotation pipeline
Timo and Hasegawa-san have appended the CAGE-Tag-Cluster-Annotation.sh script and the annotation sets mentionned ablove to their pipeline implementation, automating the annotation of HelicosCAGE clusters with respect to refseq transcript models (for all genomes) and gencode transcript models (hg19 only).
Below is the graphical representatin of this annotation pipeline :
