VAST Challnge 2015

The Visual Analytics Science and Technology (VAST) Challenge is an annual contest with the goal of advancing the field of visual analytics through competition.

• Topic: VAST, Data Contest
• Date: October 2015
• Service: Data Analysis, R

The VAST Challenge 2015 presented two related mini-challenges and an overall Grand Challenge, and we prepared submissions for each category, and our Mini-Challenge 1 received an award, Good Combination of Analysis and Visualization to Solve the Challenge.

The majority of analysis was performed in R, and all of the source code, our submissions, as well as reviewers’ comments are available from this git repository.

The following videos shows our analysis approaches and some of findings.

Gapmap

R package for gapped dendrogram and heatmap

• Topic: R, Cluster Heatmap
• Date: April 2015
• Service: Visual Analytics Tools

One of the common mistakes when interpreting a dendrogram is to assume that two adjacent nodes in the leaf order to be similar. To address this common pitfall, we developed an R package named gapmap.

For example, if you perform hierarchical clustering based on the Euclidian distance between European cities shown below, it results in the following dendrogram structure.

If you look at nodes of the dendrogram, you find “Stockholm” and “Lisbon” next each other. Just because they end up right next each other, it does not mean they are similar. We know that the distance between “Lisbon” and “Madrid” is actually smaller than that of “Stockholm” and “Lisbon”.

Because the proximity (Gestalt’s principles) is a strong visual cue, and the distance between the cities is only encoded vertically on dendrograms, it results in this common misinterpretation. To address this point, we encodes the distances between nodes by introducing gaps.

This is a rather simple idea, but it results in a much more intuitive representation of clusters of European cities. This technique can also be extended to cluster heatmap visualizations.

The naming of “gapmap” was first introduced by Nils Gehlenborg and Bang Wong in the Point of view column in Nature Methods.

We developed the method in R and published as a package gapmap on CRAN. The vignettes for the package is also available. Example 1, Example 2.

Anatomical Illustration

Comprehensive molecular characterization of gastric adenocarcinoma - The Cancer Genome Atlas Research Network

• Topic: Anatomy
• Date: September 2014
• Service: Figure Design

We worked closely and iteratively with a group of researchers from the TCGA network to prepare a figure for their publication. The above is the final figure, and the below shows the iterations to refine the anatomical illustration.

Dendsort

dendsort modular leaf ordering methods for dendrogram representations in R

• Topic: R, Hierarchical clustering
• Date: July 2014
• Service: Visual Analytics Tools

Dendrograms are graphical representations of binary tree structures resulting from agglomerative hierarchical clustering. In Life Science, a cluster heat map is a widely accepted visualization technique that utilizes the leaf order of a dendrogram to reorder the rows and columns of the data table. The derived linear order is more meaningful than a random order, because it groups similar items together. However, two consecutive items can be quite dissimilar despite proximity in the order. In addition, there are 2^n-1 possible orderings given n input elements as the orientation of clusters at each merge can be flipped without affecting the hierarchical structure. We present two modular leaf ordering methods to encode both the monotonic order in which clusters are merged and the nested cluster relationships more faithfully in the resulting dendrogram structure.

For more details, please check out our publicaiton, the CRAN repository, and its vignettes.

This is an animation demostrating what the algorithm is doing…

The idea for developing this R package came during a research visit to the Shmulevich lab at the Institute for Systems Biology in Seattle. Working closely with research scientists (Vesteinn Thorsson and Sheila Reynolds), we eceloved an interactive heatmap visualization tool as shown below.

Cosmoplitan Chicken Research Project

Data Sculptures to represent SNP array data of chickens

• Topic: Data Sculpture
• Date: April 2014
• Service: Outreach, Art

In collaboration with a Belgian artist, Koen Vanmechelen, we generated data sculptures based on the SNP array data of chickens from the Cosmopolitan Research Project.

Koen Vanmechelen started the Cosmopolitan Chicken Project to reflect on the issues of (biological, cultural, behavioural, …) diversity in the human population in 90s. Vanmechelen crosses pure-bred chicken lines to create the true “cosmopolitan chicken”. At present, 16 different breeds have been combined in a single hybrid.

Each ring represents a chromosome, thus varying lengthes. Spikes pointing inwards represents homozygous polymorphisms and those pointing outwards represents heterozygous polymorphisms. As a result, inbred chickens have more peaks pointing inwards while crossbred chickens have more peaks pointing outwards.

These sculptures were 3D printed using the Selective Laser Sintering (SLS). Vanmechelen incorporated these data-driven sculptures into his exhibitions at Biennial of Venice and BOZAR Electronic Arts Festival in 2013.

We received the Best Artwork Award at the ISMB 2013.

Brain Constellation

Analysis of resting state functional connectivity networks.

• Topic: Neuroscience, Data Contest
• Date: July 2014
• Service: Data Analysis

The BioVis 2014 data contest focused on the domain of resting state functional connectivity networks (rs-fMRI network). Our entry received the Overall Favorite Data Contest Award at BioVis2014 (ISMB 2014). This was a collaborative work with Nico Verbeeck, Jaak Simm and Jan Aerts.

Sequence Diversity Diagram

Visualization of multiple sequence alignments sets.

• Topic: BioVis, Proteomics
• Date: July 2013
• Service: Visual Analytics Tools

The sequence logo has been the long-standing convention for visualization of multiple sequence alignments of nucleotides or amino acids for the past few decades. Although it effectively communicates the amount of information / conservation present at every position, this visual encoding falls short when the analysis task is to compare between two or more sets of aligned sequences.

We participated in the BioVis 2013 redesign contest, and our design received an honorable mention.

The paper on Sequence Diversity Diagram is available from here.

This video demonstrates a visual analytics tool (called Seagull) to include circular representation the mutual information between positions and the Jmol (Java molecular viewer) to explore the highly conserved positions in 3D structures.

The below are some data sketches we made during the design process.

The original Sequence Logo figure. First prototype.

TrioVis

A visualization apporach for filtering genomic variants of parent-child trios.

• Topic: Bioinformatics, NGS
• Date: May 2013
• Service: Visual Analytics Tools

TrioVis is a visual analytics tool developed for filtering on coverage and variant frequency for genomic variants from exome sequencing of parent-child trios. In TrioVis, the variant data are organised by grouping each variant based on the laws of Mendelian inheritance. Taking three Variant Call Format (VCF) files as input, TrioVis allows the user to test different coverage thresholds (i.e. different levels of stringency), to find the optimal threshold values tailored to their hypotheses, and to gain insights into the global effects of filtering through interaction.

Publication. Git Repository.

Horner's Syndrome

Interupption of the cervical sympathetic pathway from the hypothalamus.

• Topic: Neuroanatomy
• Date: December 2008
• Service: Figure Design

This illustration was created for the Neuroanatomy module during the MSc in Biomedical Communications.

The below are two versions of the illustration.