CURRENT CHALLENGES IN GENOMIC DATA VISUALIZATION Cydney Nielsen BC - - PowerPoint PPT Presentation

CURRENT CHALLENGES IN GENOMIC DATA VISUALIZATION

Cydney Nielsen

BC Cancer Agency Genome Sciences Centre Vancouver, Canada

The Data Deluge

~$5,000 in 2001 ~10¢ in 2011

Sequencing Experiments

De novo assembly Re-sequencing Enrichment

AGCTTCAGATGGACAGATAA GGCATACAGACTTAGACATA CCAGACAAGACAGACACAGTA TACAAGACATAAGCAATACAGA CCAGACAAGACAGACACAGTA AGCTTCAGATGGACAGATAA GGCATACAGACTTAGACATA CCAGACAAGACAGACACAGTA TACAAGACATAAGCAATACAGA CCAGACAAGACAGACACAGTA AGCTTCAGATGGACAGATAA GGCATACAGACTTAGACATA CCAGACAAGACAGACACAGTA TACAAGACATAAGCAATACAGA CCAGACAAGACAGACACAGTA

Reference Genome Reference Genome Genome Assembly

Drew Sheneman, New Jersey - The Newark Star Ledger

Challenge 1

Large number of samples for comparison

“To systematically characterize the genomic changes in hundreds of tumors…and thousands of samples over the next five years” The Cancer Genome Atlas www.cancergenome.nih.gov

Genome Browsers

Stacked data tracks along a common genome x-axis

Genome coordinate Data samples

Home Genomes Blat Tables Gene Sorter PCR PDF/PS Session FAQ Help

UCSC Cancer Genomics Heatmaps

Glioblastoma Copy Number Abnormality, Agilent 244A array (n=200)

Tumor vs normal G e n d e r

Recurrent deletion of all or part of chromosome 10, peak at PTEN locus

Genome coordinate Data samples Zhu et al., Nature Methods, 2009

Heatmap provides a more condensed view

Challenge 1

Consider what information is needed

e.g. replace with biologically meaningful summary, such as significant change between samples

Large number of samples for comparison

Home Genomes Blat Tables Gene Sorter PCR PDF/PS Session FAQ Help

UCSC Cancer Genomics Heatmaps

Glioblastoma Copy Number Abnormality, Agilent 244A array (n=200)

Tumor vs normal G e n d e r

Recurrent deletion of all or part of chromosome 10, peak at PTEN locus

Zhu et al., Nature Methods, 2009

Example: Summary view (column averages)

Challenge 2

Large number of data types

Genomic location (Mb) 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 2 4 1 Copy number Allelic ratio 100 Non-inverted

• rientation

Inverted

• rientation

Deletion-type Tandem dup-type Head-to-head inverted Tail-to-tail inverted

SNU-C1 (colorectal): Chr 15

A

Stephens et al., Cell, 2011

Genomic rearrangements in cancer (complex representation)

Keane et al., Nature, 2011

CAST/EiJ WSB/EiJ PWK/PhJ

b a

SPRET/EiJ 1 2 9 P 2 / O l a H s d 1 2 9 S 1 / S v I m J 1 2 9 S 5 / S v E v

B r d

A / J A K R / J B A L B / c J C 3 H / H e J C 5 7 B L / 6 N J C B A / J D B A / 2 J L P / J N O D / S h i L t J N Z O / H I L t J

1 2 3 4 5 6 7 8 9 10 1 1 1 2 13 14 15 16 17 18 19 X X 19 18 1 7 16 15 1 4 13 12 11 10 9 8 7 6 5 4 3 2 1 X 19 18 17 16 15 1 4 13 1 2 1 1 1 9 8 7 6 5 4 3 2 1 X 1 9 1 8 1 7 16 15 14 13 12 11 1 9 8 7 6 5 4 3 2 1 X 19 18 1 7 16 15 14 13 1 2 11 1 9 8 7 6 5 4 3 2 1

>100,000 742 179 836 SNPs SVs TEs Uncallable

17 mouse genomes (more compact representation) Still difficult to represent many data types in a general tool

Challenge 2

Compact, customized data encoding Large number of data types

inversion event in a human lymphoma genome reference human genome

(a) (b)

Nielsen et al. Best Paper Award at InfoVis 2009

ABySS-Explorer

Represents sequence

• connectivity
• strand
• length
• mapping on reference

Interactively access

• sequence coverage
• scaffolding

Challenge 3

Genomic features are sparse

Genome Browsers

LOCAL VIEW

Human chr1, 1 pt corresponds to 480 kb, which is larger than 98% of all human genes!

• Martin Krzywinski

a b

Chromatin states: Chromosome 3L 5′ 3′ 5′ 3′ 5′ 3′ 5′ 3′

Pericentromeric heterochromatin Cluster of small expressed genes PcG domains Heterochromatin- like domain Open chromatin domain

9 8 7 6 5 4 3 2 1

Hilbert Curve

GLOBAL VIEW

Kharchenko et al., Nature, 2011 Anders, Bioinformatics, 2009

Challenge 3

Genomic features are sparse Need both overview and detail Functional axis (perhaps not full genome)

H3K4me3 H3K9Ac H3K4me1 H3K36me3 H3K27me3 H3K9me3 MeDIP MRE

• 2. ¡Extract ¡data ¡matrices ¡
• 1. ¡Focus ¡on ¡regions ¡of ¡interest ¡(e.g. ¡transcrip8onal ¡start ¡sites) ¡
• 3. ¡Cluster ¡matrices ¡

¡

• 4. ¡Interac8ve ¡cluster ¡visualiza8on ¡ ¡

Spark – a genomic data exploration tool

Nielsen et al. in preparation

Challenge 4

No longer one genome but many

Single nucleotide variation

Ossowski et al. Genome Research, 2008

Single nucleotide variation

Integrative Genomics Viewer (IGV)

Robinson et al. Nature Biotechnology, 2011

Bhutkar et al., Genetics, 2008

Structural variation

Challenge 4

No longer one genome but many Capture variation on a graph

Comeau et al., Mol. Biol. Evol., 2010

Sequence variation on a graph

Users may require more time to learn how to interpret graph representations, but such graphs are likely to scale better and may prove more powerful for analysis

Paten et al., Genome Research, 2011

Sequence variation on a graph

Challenge 5

Human Judgement Computational Analysis

Consed Genome Assembly and Finishing Tool

David Gordon and Phil Green Good example of integrated visualization and computational analysis functionality

Challenge 5

Need to integrate computation High interactivity, low memory overhead Avoid storing large data sets locally Popularity of web-based tools Evolving sequencing technologies

Summary

Large number of samples for comparison Large number of data types Genomic features are sparse No longer one genome but many Need to integrate computational analysis 1 2 3 4 5