Leveraging Overview For Scalable Genomic Alignment Visualization

Sequence SurveyorLeveraging Overview for ScalableGenomic Alignment VisualizationDanielle Albers, Colin Dewey, and Michael GleicherUniversity of Wisconsin-MadisonDepartment of Computer SciencesIEEE VisWeek 2011

Viewing Genome Alignments

Viewing Genome Alignments

PerceptionScalable DesignAggregationMapping

Scalable Design

OutlineThe Data DomainSequence SurveyorDesign in Theory- Perception- Mapping- AggregationDesign in Practice

Whole Genome AlignmentIdentify related groups of genesappearing in a set of organisms

Defining ScaleNumber of GenomesLength of GenomesTypes of Inquiry

OutlineThe Data DomainSequence SurveyorDesign in Theory- Perception- Mapping- AggregationDesign in Practice

Our Solution

Our SolutionBlock DetailMappingPanePhylogeneticTreeHistogramGenomes

Our SolutionPerceptionGenomes

Our SolutionBlock DetailAggregation

Our SolutionMappingPaneMapping

Our SolutionPhylogeneticTreeHistogram

OutlineThe Data DomainSequence SurveyorDesign in Theory- Perception- Mapping- AggregationDesign in Practice

PerceptionHow the user processes dense dataInform scalable design- Limitations of current designs- Insight into future designsFour principles

Perceptual PrinciplesPre-Attentive PhenomenaVisual ClutterVisual SearchSummarization

Perceptual PrinciplesPre-Attentive PhenomenaVisual ClutterVisual SearchSummarization

Perceptual PrinciplesPre-Attentive PhenomenaVisual ClutterVisual SearchSummarization

Perceptual PrinciplesPre-Attentive PhenomenaVisual ClutterVisual SearchSummarization

Perceptual PrinciplesPre-Attentive PhenomenaVisual ClutterVisual SearchSummarization

PerceptionOverview - Sacrifice detail for high-levelcomparisonColorfield - Emphasize visual structureMappings – Emphasize key detailsAggregation – Do not overwhelm viewers

MappingColor MappingColor SchemesPosition Mapping

Combinations of different color and positionmappings reveal interesting trends in the dataPos in ReferenceGrouped FreqIndexIndexMembership FreqGrouped FreqPos in Reference

AggregationCannot show all the data at once- Limited screen real estate- ClutterBlocking preserves local control- Display gene neighborhoods as glyphsFour block encodings

BlockingGroup (relatively) continuous sets ofneighboring genes into a single unittilSrofyaeQphnAtadG

Aggregate EncodingsAverage

Aggregate EncodingsAverageRobust AverageColor WeavingEvent Striping

InteractionBlock Brushing: Highlight locations of block contentsin overview, phylogeny, and histogram onmouse-overBlock Linking: Link locations of block contents inoverview on clickDetail Notes: Details of genes in a block andmatching genes of the set are presented in aseparate windowNon-locality Zoom: Explore the contents of anaggregate block in the Block Detail Window onmouse-overZoom Lock: Fix the contents of a block in the zoomwindow to explore the distributions of specificgenesManual Rearrangement: Drag-and-droprearrangement of sequences and indicatebranch crossings by opacityFiltering: Highlight genes matching a set of names, idnumbers, frequencies, genomes, or chromosomesLoad Filter: Load a filter set from a CSVSave Filter: Save the current filter set to a CSVHistogram Brushing: Highlight the locations of genes ina region of the frequency distribution in theoverview and phylogenetic tree by mouse-overZoomed Gene Brushing: Highlight locations of genesin overview, phylogeny, and histogramLoad Tree: Load different trees and arrangements froma tree fileZoomed Gene Linking: Link locations of a set ofmatching genes in the overviewSave Tree: Save the current tree structure andsequence arrangement to a tree file

OutlineThe Data DomainSequence SurveyorDesign in Theory- Perception- Mapping- AggregationDesign in Practice

Use Cases100 Bacteria6,000 genes50 Bacteria5,000 genes35 Fungi17,000 genes14 Pathogens4,000 genes8 partial E. coli sequences300 genes

ParallelsCan use Sequence Surveyor to obtaininformation presented in existing toolsat scale.Mauve: Color by position in reference (arrow), order by start position

Anecdotes: BuchneraBuchnerafamily ofgenomes andthe ancestralcoreColor by position in reference (arrow), order by set of genomes containing each gene

Anecdotes: BuchneraAveraging:No significant trendColor Weaving:Overall distribution

Anecdotes: E. ColiConservation relationships between different families of genomesColor by position in reference (arrow), order by relative ordering

Anecdotes: FungiBioinformatics applications allow users to test algorithms using visual checksColor by overall frequency, order by relative ordering

Anecdotes: FungiBioinformatics applications allow users to test algorithms using visual checksColor by position in a reference, order by relative ordering

ExtensionsProteins andnucleotide MSAAny data with anorthology andordered setsTop 5,000 most popular words since 1660Google N-GramsDistribution of a word set in 2000 across time

SummaryScalable whole genome alignment overviewPerception informs designUser-controlled mapping scales across queriesAggregation filters dataExtends beyond the immediate biology

AcknowledgementsUniversity of Wisconsin – MadisonDepartment of ComputerSciences Graphics & Vision LabUniversity of Wisconsin – MadisonBACTER Institute for ComputationalBiologyUniversity of Wisconsin – MadisonGenome Center Genome EvolutionLaboratoryDr. David BaumlerDr. Eric Neeno-EckwallDr. Jeremy GlasnerDr. Nicole PernaFunding by NSF awards IIS-0946598, CMMI-0941013 and DEB-0936214 andDoE Genomics: GTL and SciDAC Programs (DE-FG02-04ER25627)

AvailabilityPrototype and sample data package (coming eyor/dalbers@cs.wisc.edu

in overview, phylogeny, and histogram Zoomed Gene Linking: Link locations of a set of matching genes in the overview Manual Rearrangement: Drag-and-drop rearrangement of sequences and indicate branch crossings by opacity Filtering: Highlight genes matching a set of names, id numbers, frequencies, genomes, or chromosomes