Di 2011 houston

21 22
1
20

50M

50M
0M

0M

100M
0M

M
50M

150
19

M
200
0M
M
50

0M
0M
18 2

M
50
M
50

0M
ggbio
10
0M
0M
15
l 2
17

0M
20
M
50
l l

0M
l strand
0M
l
+
16

M
50M 50

−

3
l
M
0M 100
l

100M l
150
M rearrangements
1
15

50M
l
l
interchromosomal
l l
0M
l
l
0M
intrachromosomal

50M
100M tumreads
48245000 48250000 48255000 48260000 48265000 48270000

4
l
4
14

l
100M
hg19::chrX
50M

l 6

Extending the
0M 150M

l
l 8
l
100M
0M l 10
13

l
l 12
50M
50M
0M l

100

5
M
l
15
0M
10 l
0 M

50 l
12

0M
M

0M
50
M
l 10
0M
10

6

Grammar of Graphics
0M

15
50

0M
M

11
0M

0M
50
10

M
100
0M

50M

7
150

M
0M

0M

10
M
50M
100M

100M
50M

0M

9 8

to Genomic Data
1

Tengfei Yin, Di Cook
2
3
4
5

Stepping
6

Iowa State University
7
8
9

gieStain
10

acen

Michael Lawrence
gneg
11

gpos100
12

gpos25
13

gpos50
gpos75
14

gvar
15

15
stalk

Genentech
16
17

Coverage
18

10
19
20
21

5

Interface 2012, Rice University
22
X

0
Y

0.0e+00 5.0e+07 1.0e+08 1.5e+08 2.0e+08 0 100 200 300

Motivation
Lots of tools exist for displaying
genomic data
Many different packages, many
standalone, many different data
standards

ggbio - Genomic Data Vis - Interface 2012, Rice University 2 /31

Motivation


Motivation
Circos


Motivation
Circos
Need construct a central and many other
configuration files from scratch, learning
curve is very high
Adding legend not easy
Cannot map aesthetics to certain
variables


Motivation


Motivation
UCSC Genome Browser


Motivation
UCSC Genome Browser
Karyogram view,
with associated
data plotted


Motivation
UCSC Genome Browser

Logical zoom, all
we know about
this genetic code


Motivation
UCSC Genome Browser
Very commonly used, very popular
Gives broadly applicable, generic, but
narrow selection of plot choices
No operations on genomic ranges views to
facilitate perception of structure


Motivation
Chromosome X

65.93 mb 65.95 mb 65.97 mb

65.94 mb 65.96 mb

70
60
50
40
30
4
2
0
−2

5 Composite plots for multiple chromosomes

Motivation
Chromosome X
Gviz (Hahne et al)
65.93 mb 65.95 mb 65.97 mb

65.94 mb 65.96 mb

70
60
50
40
30
4
2
0
−2

5 Composite plots for multiple chromosomes

Motivation
Gviz (Hahne et al)


Motivation
Gviz (Hahne et al)
Pretty good!
Incorporated with R, and R data
structures
Uses grid (low level) graphics, very
flexible, but not leveraging tools like
ggplot2


Outline
What is the grammar of graphics?
How it is extended for genomic data.
Examples
Next steps: interactive graphics


Grammar
Grammar forms the foundation of a
language. It is a set of structural
rules that govern composition.
For graphics, it provides a way to
construct a plot in a common form,
and enables clarification of
similarities and differences between
plots.


Grammar (ggplot2)
80

Bar chart 60 day
Thu
ggplot(data=tips,

count
40 Fri
Sat

aes(x=day, fill=day)) + 20
Sun

geom_bar(width=1) 0
Thu Fri Sat Sun
day


Grammar (ggplot2)
80

Bar chart 60 day
Thu
ggplot(data=tips,

count
40 Fri
Sat

Sun

geom_bar(width=1) 0
Thu Fri Sat Sun
day

Pie chart
ggplot(data=tips,
aes(x=day, fill=day)) +
geom_bar(width=1) +
coord_polar()

Grammar (ggplot2)
80

Bar chart 60 day
Thu
ggplot(data=tips,

count
40 Fri
Sat

Sun

geom_bar(width=1) 0
Thu Fri Sat Sun
day

Pie chart 80
Sun Thu
60
ggplot(data=tips, 40
20
day
Thu

count
0 Fri

aes(x=day, fill=day)) + Sat
Sun
Sat Fri
geom_bar(width=1) +
coord_polar() day


Grammar (ggplot2)
80

Bar chart 60 day
Thu
ggplot(data=tips,

count
40 Fri
Sat

Sun

geom_bar(width=1) 0
Thu Fri Sat Sun
day

80
Sun Thu
60 day
40
20 Thu

count
0 Fri
Sat
Sun
Sat Fri

day


Grammar (ggplot2)
80

Bar chart 60 day
Thu
ggplot(data=tips,

count
40 Fri
Sat

Sun

geom_bar(width=1) 0
Thu Fri Sat Sun
day

Rose plot/Coxcomb 80
Sun Thu
60
20
day
Thu

count
0 Fri

aes(x=day, fill=day)) + Sat
Sun
Sat Fri
geom_bar(width=1) +
coord_polar() day


Grammar (ggplot2)
Stacked bar chart 200

day
Thu

count
aes(x=””, fill=day)) + 100
Fri
Sat
Sun

geom_bar(width=1) 50

0

""


Grammar (ggplot2)

day
Thu

count
Fri
Sat
Sun


0

""

Pie chart
ggplot(data=tips,
aes(x=””, fill=day)) +
geom_bar(width=1) +
coord_polar(theta="y")

Grammar (ggplot2)

day
Thu

count
Fri
Sat
Sun


0

""

Pie chart 0

50
day
Thu

count
Fri

aes(x=””, fill=day)) + Sat
Sun

geom_bar(width=1) + 150
100

coord_polar(theta="y") ""


Grammar Elements
DATA: What is to be plotted
STAT: Statistical operations to make on
data, like binning.
GEOM: Geometric object, elements to use
to displays aspects of the data
SCALE: Map data to aesthetics to geom
COORD: Coordinate system to use, eg
Cartesian
(FACET): subset and display

Example: MA plot


Example: MA plot

DATA=expression data frame,
x=average - Genomic Data Vis - Interfacechange University
ggbio intensity, y=fold 2012, Rice 11 /31

Example: MA plot
GEOM=point


Example: MA plot

SCALE=x is
logged


Example: MA plot

SCALE=color is
mapped to
statistical
significance


Example: MA plot

COORD=default,
Cartesian


Example: MA plot

FACET=none


Example: MA plot
qplot(baseMean, log2FoldChange,
data = res, geom = "point",
xlab = "Normalized mean",
ylab = "log2 fold change",
xlim = c(0, 10000),
color = group) +
scale_x_log10() +
scale_color_manual(
values = c("black", "red"))

What’s different?
Genomic data has interval context
Several common geoms used in
standard plots, not in current
grammar
Additional transformations common
Lining up of multiple data plots,
especially against genome


What’s different?
No seqnames ranges strand tx id exon id
1 chrX [48242968, 48243005] + 35775 132624
2 chrX [48243475, 48243563] + 35775 132625
3 chrX [48244003, 48244117] + 35775 132626
4 chrX [48244794, 48244889] + 35775 132627
5 chrX [48246753, 48246802] + 35775 132628
... ... ... ... ... ... ...
26 chrX [48270193, 48270307] - 35778 132637
27 chrX [48269421, 48269516] - 35778 132636
28 chrX [48267508, 48267557] - 35778 132635
29 chrX [48262894, 48262998] - 35778 132633
30 chrX [48261524, 48262111] - 35778 132632

able 2: Typical biological data coerced into a data frame: A GRanges table representing gene SSX4 an
DATA: Genomic ranges
SX4B. One row represents one exon, seqnames indicates the chromosome name, ranges indicates the interva
exons, strand is the direction, tx id and exon id are the internal id’s used for mapping cross database.


Extensions
data source(s)
I/O packages in bioconductor

abstract data (formal model) meta data

autoplot geom stat scale grammar of
graphics with
coord facet layout extensions

plots tracks


Extensions
current software. Development of new visualization tools should be
independently factorized into components of the grammar. Table 1
describes the extensions developed in this work.

Comp name usage icon

geom geom rect rectangle

geom segment segment
aut
geom chevron chevron

geom arrow arrow

geom arch arches

geom bar bar

geom alignment alignment (gene)

stat stat coverage coverage (of reads) Figu
ggbio - Genomic Data Vis -mismatch pileup forRice University
Interface 2012, tion.
16 /31
stat mismatch mod

geom arrow arrow

Extensions
geom arch arches

geom bar bar

geom alignment alignment (gene)

stat stat coverage coverage (of reads) Figure
tion. It
stat mismatch mismatch pileup for model,
alignments map d
new co
stat aggregate aggregate in sliding
body o
window
ggbio.
stat stepping avoid overplotting

stat gene consider gene struc-
ture
Sev
In ggp
stat table tabulate ranges ple bin
resenti
stat identity no change comm
covera
coord linear ggplot2 linear but
ggbio - Genomic Data Vis - Interface 2012, Rice University
facet by chromo- 17 31al
read
/
Add

stat gene consider gene struc-
ture
Several

Extensions
In ggplot2
stat table tabulate ranges ple binning
resenting a
stat identity no change commonly
coverage,
coord linear ggplot2 linear but
facet by chromo-
read alignm
some Additio
faceting m
genome put everything on are listed in
genominc coordi- examples.
nates
Let’s an
truncate gaps compact view by Figure 3, t
shrinking gaps In this plo
by using th
layout track stacked tracks attributes s
system. It

chr1
chr2
karyogram karyogram display
ing genom

chr3
50 100 150 200 250 300
start

circle circular the differe
the gramm
faceting formula facet by formula once one g
discover th
ranges facet by ranges componen
scale not extended ggplot2default in the desig
ggbio - Genomic Data Vis - Interface 2012, Rice University /
18 31
The foll

Extensions
autoplot
Tries, and does a jolly good job, of
recognizing the data object to be
plotted, and how it should be
displayed.


Example

2

strand
+
−

1

48245000 48250000 48255000 48260000 48265000 48270000
hg19::chrX


Example

2

strand
+
−

1

48245000 48250000 48255000 48260000 48265000 48270000
hg19::chrX

DATA=GRangesList Object


Example
GEOM=alignment, chevron
2

strand
+
−

1

48245000 48250000 48255000 48260000 48265000 48270000
hg19::chrX


Example

2

SCALE=stepping, strand
+

color=strand
−

1

48245000 48250000 48255000 48260000 48265000 48270000
hg19::chrX


Example

2

strand
+
−

1

48245000 48250000 48255000 48260000
hg19::chrX LAYOUT=linear
48265000 48270000


Examples
p1 <- autoplot(gr)
p2 <- autoplot(gr,
Stepping

stat = "coverage")
tracks(p1, p2)
Examine short
15
reads
Stack them (top)
Coverage

10

5 Collapse into
0
“density” (bottom)
0 100 200 300


Examples
uc002dwc.3(226)
p1 <- autoplot(txdb,
which = genesymbol["A"])
uc010veg.2(226)

uc002dwa.4(226)
which = genesymbol["A"],
uc002dvz.3(226)
stat = "reduce")
uc002dvw.3(226)
tracks(p1, p2,
heights = c(4, 1))
uc002dvx.3(226)

Compare transcripts
uc010bzo.2(226)

Reduce all to one
30065000 30070000 30075000 30080000


Examples
uc002dwc.3(226)
uc010veg.2(226)

uc002dwa.4(226)
which = genesymbol["A"])
knownGene

uc002dvz.3(226)
which = genesymbol["A"],
uc002dvw.3(226)

uc002dvx.3(226)

uc010bzo.2(226)
truncate.gaps = TRUE)
30065000 30070000 30075000 30080000
hg19::chr16
tracks(p1, p2,
uc002dwc.3(226)

uc010veg.2(226)
heights = c(4, 4))
uc002dwa.4(226)
knownGene

uc002dvz.3(226)

uc002dvw.3(226)

uc002dvx.3(226)

uc010bzo.2(226)

30064411
hg19::chr16
30081741
Focus on exons


Di 2011 houston

Recommended

Recommended

More Related Content

Featured

Featured (20)

Di 2011 houston