Finalize Shiny prototype layout

[jaeddy]

The Shiny app will initially be the initial version of the iAtlas portal on release (pre-April 5th, 2018). Need to settle on a more suitable overall layout for visualization panels and any other information we want to include.

[andrewelamb]

The development version is now in dashboard form, is this sufficient?

[jaeddy]

Definitely a good start, but still some more figuring out to do in terms of organization and layout. I'll take a pass on the current version.

[jaeddy]

Feeling OK about overall layout after PR #22; will attempt to tackle module organization next. After looking at the manuscript figures some more, I think the modules we have are a reasonable starting point.

Note: I'm envisioning the modules as all being fairly "rich" with multiple sections and views to explore results related to a particular theme (this is inspired somewhat by the ICGC Data Analysis modules, which each have multiple components).

My current idea would look something like this:

Sample Groups Overview:

• high level trends across sample groups — e.g., Figure 1C (~ "Feature Trends"), 1D (~ "Cancer Types"), 3A ("Overall Survival")
• extends what we currently have for the "Feature Correlations" module

Tumor Composition

• some plots from Figure 1 would apply, most of Figure 2, and possibly parts of S3
• extends the current "Cell Content" module
• idea would be to present trends of "Overall Cell Proportions" (e.g., stromal, LF, etc.) and "Cell Type Fractions" (e.g., CIBERSORT, aggregates) across sample groups
• I think it would also make sense to include prognostic trends (concordance index) as a function of immune cell type (e.g., S3C, S3D)

Clinical Outcomes

• "Overall Survival"
• "Concordance Index"
• etc.

Genomic State

• covers most of Figure 4 (not sure about S4)
• not currently represented with existing modules
• summarizes trends like "Mutational Load", "Intratumor Heterogeneity", "Aneuploidy", and "Neoantigens" across sample groups
• this particular module might be more complicated / tricky to put together in a logical way...

Immune Interface

• basically everything in Figure 5, optionally parts of S5 as well
• extends the current "Clonal Diversity" module
• topics/sections would include ~ "Peptide-MHC Distribution", ~ "Adaptive Immune Repertoire Diversity", etc.

Immunomodulator Genes

• for lack of a better name... pretty much Figure 6/S6
• extends the current "Immunomodulators" module
• presumably focusing on expression for now, but can include CNV, methylation info if we have time

Regulatory Networks

• all the networks! (Figure 7)
• more ambitious / less likely to get done for our 4/5 target

Taking a stab at mocking up these ideas in the app today! Comments welcome.

Other thoughts:

• I'm not sure if it makes sense to have a separate module for prognostics / survival, as a lot of these results are related to other categories and features — I could be convinced otherwise.
• I'm also contemplating whether to make "sample groups" a global option — i.e., users would select 'Immune Subytpe', 'TCGA Study', 'TCGA Subtype' in the left/side panel and results for each module would be broken down accordingly.

[Gibbsdavidl]

On Mar 26, 2018, at 9:09 AM, James Eddy notifications@github.com wrote:

I'm not sure if it makes sense to have a separate module for prognostics / survival, as a lot of these results are related to other categories and features — I could be convinced otherwise. I think it makes sense, especially if people go: choose a sample grouping (or upload one? maybe for later) then see the survival curves.

I'm also contemplating whether to make "sample groups" a global option — i.e., users would select 'Immune Subytpe', 'TCGA Study', 'TCGA Subtype' in the left/side panel and results for each module would be broken down accordingly.

Yes, I like it. Because additionally, I like making sample groups out of features like has/or-has-not a given mutation in gene X. Then you might look at the survival curves. If we could (even very simply) define sample groups from our features, that would be really cool. Maybe save for later.

-dave

[jaeddy]

OK, makes sense! Will plan to keep survival as a separate module for now.

And yeah — making the sample groups a global option would definitely enable one to "define" their own groups somehow. This is also more similar to how the ICGC Portal works. I think for now, we can just require people to use our pre-defined groups. :)

[jaeddy]

After speaking with @vthorsson, we're going to hold off on including the Genomic State and Immune Interface modules for the initial release. For the latter, the information we currently present in clonaldiversitymodule should be covered by the violin plots in Sample Groups Overview.

I'll comment these sections out for now, so we can focus our attention more easily.

I've been sketching out ideas for possible layout of the Sample Groups Overview and Tumor Composition pages, which I'll test out in the app soon.

[jaeddy]

Calling this done for now (as of PR #34)