caleydo / vials

Vials is a Caleydo Web application for visualizing alternative splicing based on mRNAseq data.

vials

Vials visualizes alternative splicing based on mRNAseq data. You can try out vials at http://vials.io/.

Installation

git clone https://github.com/caleydo/vials.git
cd vials
npm install

Testing

npm test

Building

npm run build

---

This repository is part of Phovea, a platform for developing web-based visualization applications. For tutorials, API docs, and more information about the build and deployment process, see the documentation page.

vials's Issues

Collapse individual groups into one Categorical Column

Here is a suggestion of how to collapse the group into one categorical column:

Show introns when using broken scales between Isoforms as squiggly lines

Similar to what ABySS Explorer does
http://www.bcgsc.ca/platform/bioinfo/software/abyss-explorer

The frequency of the oscillations should correspond to the length of the intron

Prepare for read counts instead of direct read information for exons

For TCGA data we will get read counts instead of reads for exon abundance.

Allow comparison of how conditions behave across junctions of an isoform

In this image on the left, we see the default state on the left, and a new state on the right:

The state on the right shows a "fanned out version" where only the junctions of one specific isoform is visible. (this could also be part of the details junction view instead of the overview junction view).

The points are sorted in x by their y value for the selected junction, for all other junction the sort order of the selected junction is preserved. This will make it easy to spot any deviating patterns of conditions in junction reads.

Add spacing between junction UI elements

There should be a spacing on the left and between all elements.

Fix initialization of crosshair

Not really a problem, but if it's not a lot of effort I think it would look better.

fix intron toggle functionality

Fix consistency fo gene data

Curently, the AltSpliceGUI module injects specific gene data into the other modules and shows the parameters (chromosome, start position, base width) in the header as non-editable information (Chr 17, 45814875, 1500).

The two JSON files provided by Joseph
pileup ENSG00000149557
pileup ENSG00000150782
contain jxn weights between Exons not within the above range mandated by AltSpliceGUI, there are Exons for example at position 112014540 and 125316007 in the above files.

This makes the "altspice-junctions" module unable to match the weights to the exons, and fails to create the visualization (with a lot of error messages).

How should we fix this? I can temporarily ignore the gene information injected by AltSpliceGUI and infer the actual exon locations from the data file itself, which might makes the jxn modlue work, but this does not solve the problem (esp. w.r.t to coordination between the views).

Background of Junction View disappeared

label in reads view

truncate the string at a maximum length.
if string is longer than threshold replace last 2 chars with '...'

Collapsing rows based on categories

Similar to what we do right now with collapsing based on groups, we want to collapse based on general categories (see #31). If we have multiple categorical dimensions, we will only be able to collapse based on one of these at a time. This figure shows the collapsing of the rows based on the gender dimensions:

Integrate isoform and isoform frequency view in the tool

Is still in different pieces of code afaik.

Normalize junction frequency to exon frequency?

One comment we got in our discussion was whether we could somehow normalize the junction frequency based on the exon frequency. This does make sense - you'd expect more junctions if both exons involved have more reads. However, there are some challenges:

• it's not clear to what we should normalize - average reads per base across both exons? average reads of the less frequent exon?
• this is an additional abstraction that users have to understand and we have to communicate
• the design is not obvious

We probably could switch the junction frequency view between the two options on-demand.

Implement sample comparison for junctions

One of the comments by the domain experts was that it would be important to compare the junction reads across groups of conditions. I suggest to use colors in the view showing all junctions, but to use a spatial separation (also colored) for each condition in the details view. This way we could show independent boxplots and reads for hundreds of samples and 5-10 conditions efficiently.

Change junction view from exon-specific to doner/acceptor sites

Our initial understanding of the data was wrong. Exons are not just permuted, exons on different isoforms can occupy the same region but begin and end at different positions. Therefore we have to adapt our junction view. Here is a sketch:

The sketch contains three isoform examples below. The donor and acceptor sites are indicated by triangles. Donor sites show up-pointing triangles, acceptor sites down-pointing triangles.

The idea is that we now have one "box" per donor site (i.e., the end of an exon), and within that box we show all the connections to the acceptor sites. In contrast to our current design we don't show all possible acceptor sites, just those where we actually have data, i.e., they can be of varying width. And we don't keep a column free for the donor.

Within the boxes we have a miniature version of the sequence.

When an isoform is selected we show its exons on the sequence belwo and proceed as we do now, fading all others.

adjust data provider to use pre-built project structures

Coordination between Isoform View and Jxn View

1. Enable selecting an isoform in the Isoform View, to be examined in the Jxn view.
   Entry point is selectIsoform() method in Jxn view.
2. Enable deselection of an isoform
3. Enable comparision between two isoforms in the jxn view by allowing selecting a
   max of 2 isoforms in the isoform view

allow for user-selected sample collections

Provide larger dataset for testing purposes

Right now we only have the default sashimi use-case. We need a larger one very very soon, both in the #samples and in the #exons and junctions.

Comparison of isoform frequencies between groups of conditions

Similar to #13 we should allow a spatial separation for groups in the isoform view. This will make it possible to deal with larger number of conditions and groups.

Color code the groupings of samples

I suggest we use the 10-class paired color scale from ColorBrewer.org and pick the saturated colors.

We should draw all the lines/areas in the track view and the dots in the junction view in these colors.

We decided that for more than five groups we will switch to a non-colored version and color by selection.

Here are the colors:

a6cee3

1f78b4

b2df8a

33a02c

fb9a99

e31a1c

fdbf6f

ff7f00

cab2d6

6a3d9a

Unify design of junction view and isoform view

Specifically, we should:

• reduce height of isoform bars in junction view to the height in the isoform view (global variable?)
• change color to the one used in the isoform view (global variable in sass)
• change the colors of the isoform representation in the edge view in the same way.

Make Junction view update-able

currently it breaks when changing the dataset

add one std area spread to agg read lines

Implement box plot in isoform view

Similar to what we have in the junction view

Reduce height of junction view by 1/3

I think it's too high and would work equally well if we give it less height.

Fix initialization state of junction view

It looks like the jnct view is initialized as no groups, show all dots, no jitter. This state is not accurately reflected in the checkboxes.

Also, I suggest the following default state:

• Show all dots on
• Jitter dots on

Extend lines of boxplots across bacground area

These could be stippled, or light solid lines:

Set-up auto-deploy on Heroku

We want to auto-deploy every commit on the master to heroku so we can test easily and send links to collaborators. Here are examples for other Caleydo Web Projects:

https://caleydo-web.herokuapp.com/

Implement the details-only view for the junctions

Implement the junctions details view per the design. The details only view should transition out of the overview (i.e., it should be the same view, just with updated positions and faded-out elements).

The transition should fade all edges not part of the isoform, then reposition these junctions, then add the arcs.

If a different isoform is selected we should consider going back to the full view first to pull out the new isoforms.

Redesign how grouping is activated

This feature can't be discovered at the moment. Following suggestion: Add radios with three options:

Junction View (i)

• Compact
• Separate Groups
• Show Correlations

When hovering over the Junction View title it should display "You must select an isoform for separating groups or showing correlations"

It should not be necessary to click anything in the junction view.

For the scatterplot version, the first view should be selected by default.

Visualize Categorical Attributes of Samples

Categorical Attributes should be treated identical to the manually defined groups. We will use colored cells to visualize category memberships, or histograms when groups are collapsed:

We want to show the actual classes on mouse-over, but we should also figure out a legend for these.

Enable Group comparison in isoform view

We should allow group comparison in the isoform view. I suggest two levels:

• Color coding by groups
• spatial separation for selected groups, on demand, as drawn here:
  

The latter could use the same backbone as we have it in the jnct view right now. We could also use the same features: box plots, x-separation, etc.

"Flags" are off for TP 53/BodyMap

I assume that's because of the reverse strand?

handle exon overlap in bodymap data

For most genes in the larger data set, multiple exons are present that cover the same span (usually off by only a few base pairs)—we should handle this by merging them into a single exon view for the viz and talk with Novartis about the biological basis for doing so.

add y-axes to read view

Add Condition Labels to the Reads

In the read view, each row should correspond to a condition (heartKOa, heartKOb, heartWT1, heartWT2).

Add these labels to the beginning of the row (check https://drive.google.com/file/d/0B-zptERh1x3rNTQ1RllrSVg3WXc/view?usp=sharing).

Improve Grouping of samples

Here is a sketch to improve the grouping of the samples:

On the far left we see the state with no grouping and the first two rows are selected. Once the mouse is released, a new column for the group is created and the cells for the selected rows are filled. The other cells still have the circular selection widget.

When creating a new group, a new column is created, as shown in state 3.

When clicking a single selection widget inside a group, or when using the brush operation, these rows are added to the group.

Duplicate group membership is allowed.

Make configuration bar a single line

There aren't that many UI elements here.

• make project selection and gene selection 200 px each
• make chromosome/start position/base width 150 px each
• remove sort iso (this makes more sense as direct manipulation and works just as well)

When showing grouping in junction view, treat all ungrouped elements as one group or hide them.

We can either:

• show all not-grouped elements as one group
• hide all non-grouped elements

Add dark background of selected isoform in junction view

Right now it's only light gray, but I don't see why?

Create JSON Snapshots of the data so that we can develop the client without the server component

While our final tool will have to work with a server, we should consider producing JSON files that the client can read, equal to the server response, so that we can easily test, develop and deploy the client, independent of a server component.

Allow to switch scales to log

remove bilal_del.css

after deadline

Add scale to isoform abundance view

I suggest we do that at the bottom.

Clarify how/whether we get isoform abundace

In our meeting on Thursday it was not clear whether we have an obvious measure for a predicted isoform abundance. It seems like MISO is mainly about predicting exon frequency, but should also have some measeure of general isoform abundance. In our design we rely on isoform abundance and this was also judged as critical by the domain experts.

It seems like the Psi value of Miso isn't quite the right thing though.

Enable Group comparison in Junction View

Similar to #28 we should allow group comparison in the Junction View.

synchronize axis across views

Integrate TCGA Data

TCGA could be our second use-case. Here is more information on TCGA data:

https://wiki.nci.nih.gov/display/TCGA/RNASeq+Version+2

We won't have read-level data, and probably we won't have to run MISO ourselves for this data.

Sorting/Filtering isoform view

We should be able to sort the isoform by:

• overall average abundance
• abundance of a group
• abundance of a sample
• containment of an exon, tie-breaking by abundance

The domain experts explained that they usually want to understand all isoforms related to one exon. They typically are guided to explore isoforms through an abnormal behavior in one exon. They want to see all isoforms that contain that exon.

Consequently, we should make it easy to see all isoforms containing a specific exon. I would suggest to re-sort the isoform view to have those of the selected (one or multiple) exons on top. Break ties by average abundance of the isoform.