ucscxena / ucsc-xena-client Goto Github PK

The "Loading" jumbotron needs to disappear at some point in the event expected group data is unable to be fetched (e.g. down network connection for user, or error returned from backend during Ajax calls).

turning it off doesn't do anything

Needs to update to reflect the new hub selection.

Probably damage from the zoom animation commit. I expect a prop is not being passed correctly. Should compare to the zoom prior to the animation commit & if it's correct there, compare the inputs to plotMutationVector::drawBackground.

Cohort: TCGA Esophageal Cancer (ESCA)
Dataset: somatic mutation SNPs and small INDELs (ucsc automated vcf)
Gene: TP53

Specifically with the parameters from the image below, getting a "red-screen-of-death" screen for KmPlot modal.

To reproduce:
TCGA BRCA
Add gene expression (illuminahiseq), TP53
Filter samples to those in gene expression (agilent)
Add gene expression (illuminahiseq), TP53 (same as before)
Now filter samples to those in gene expression (illuminahiseq)
Observe grey data

Filtering by dataset in hub that's no longer there doesn't update when the hub is turned off and the refresh button clicked.
To reproduce:

1. Start up your local hub with a dataset in a cohort that is also on the main UCSC hub
2. Add a couple random columns of data to the view
3. Filter by dataset from your local hub and observe the change in the view
4. Shut down the hub and click the new refresh button
   --> Observe that data still filtered by dataset from local hub that isn't there

Also, might be nice to have slightly better CSS. It's not terrible, but it is using up a lot of extra white space.

row.map is not a function

the current "alt-click to unfreeze" requires alt-click on the visual map itself. I often try to alt-click on the tooltip itself.

If the columns are small and you click the caret menu, it displays under the column to the right. Not a big issue, but am recording here just so it's noted.

Right now it's listed at the bottom. Put it at the top instead

Example: ICGC Thyroid cancer, positional mutation, TP53. If you hover over both of the returned mutations, you can see the information is getting cut off.

To reproduce:

1. Start up your local hub with a dataset in a cohort that is also on the main UCSC hub
2. Add a column of data from that dataset to the view
3. Go to chart mode
4. Shut down your local hub
5. Hit the new refresh button (not the browser one, but the one next to the cohort)
   --> see that the column is not deleted from view

Otherwise people can change it but then they're still viewing the old data in the chart

If you add 'TP53, a', it won't render TP53.

A split second after adding a column for the following parameters will yield the red error screen below:
Cohort: "TCGA Esophageal Cancer"
Dataset: "somatic mutation SNPs and small INDELs (ucsc automated vcf)"
Gene: 'xxx'

Right now only separates genes names on commas. Need to be able to separate on space and carriage return.

Although I have incorporated the latest changes from dev branch of acthp/ucsc-xena/client, KmPlot.js does not appear to be getting the "title" prop variable. The file js/controllers/ui.js has in fact been updated within latest change:

'km-open': (state, id) => _.assocInAll(state,
['km', 'id'], id,
['km', 'title'], _.getIn(state, ['columns', id, 'columnLabel', 'user']),
['km', 'label'], Grouped by ${_.getIn(state, ['columns', id, 'fieldLabel', 'user'])}),

Sometimes, the last unit of measure shown on the x-axis partially cuts off.

There are too many significant digits.

cohort: any tcga cohort
dataset: exon expression data
example gene: ATRX

currently probes are sorted along genomic location 5'->3', not along transcript 5'->3'

Check mutation, phenotype, gene/identifier matrix

Per-column would be most flexible. Otherwise, hide globally.

It should be at the top of the list of datasets from that hub but instead it looks like it's appearing in alphabetical order

The horizontal line on the x-axis is non-existent. Only the measuring units display.

Given the selected input parameters for the Kaplar Meier graph, when hovering over the legend keys, or the actual graph lines, both the lines and legend keys will not highlight until almost 1 sec later.

Cohort: "TCGA PANCAN12"
Dataset: "PhenoTypes"
View: "_primary_disease"

Add mutation column, reload, tooltip -> throws. Could be undefined -> null conversion in json.

within a single gene, probes are arranged from 5'->3' along genomic coordinates.

Should be 5'->3' for a gene on + strand, and 3'->5' for a gene on - strand

ALK is on - strand. Typical ployA+ selected RNAseq results have 3' bias (i.e. 3' exons has higher/redder expression then 5' exons), here we see the opposite

A click w/o drag causes an error due to uninitialized variable newPos.

Should take us to dataset details.

Only tested on dense matrix. Repo by adding a column & mousing over the new column before the data arrives. Need to test mutation tracks. Problem is in HeatmapColumn.tooltip, which tries to destructure data before it is set. Need to either prevent tooltip, or display 'loading' in tooltip before data arrives.

Might also want to make the rx streams restart on error, so in the future bugs on the stream will not kill the tooltip for the remainder of the session.

Right now it's hard to know a) it's there and b) click on it.

I think we should remove groups that is empty. This will come into
critical play when KM and filter is linked. -- this is the same problem as
there are empty groups in data column before filtering.

attached data for tcga pan-cancer cohort, load to your local xena, go to tcga pan-cancer cohort, filter on this dataset, then do KM, will cause problem
Archive.zip

As on master, add text

"(Limited drawing to " + MAX + " categories)"

or
"(gene-level average)"

when appropriate

The query is updated from the one on master branch, but is returning wrong values if the probe is associated with multiple genes. Probably a server bug, but noting here so we can revert the query if necessary.

Need some sort of feedback to the user when zooming, because it's very easy now to zoom, e.g. when aiming for the resize handle.