error creating multi gene display for specific dataset

[carlocolantuoni]

hi guys,

i dont make many multi-gene displays, so i dont know if this is a general problem or specific to this dataset, but it appears i cant make any multigene displays for this dataset without getting errors (this is being done under user "Alexandre Jourdon").

i get "Could not create plot for this dataset and parameters. Please contact the gEAR team." when i hit the "plot" button.

i have tried many different variations with different metadata etc, bt alwasy get the same error with no display coming up.

https://nemoanalytics.org/multigene_curator.html?dataset_id=b3816674-c644-4eb9-843a-e9cce381e937

[adkinsrs]

Taking a look at this today

[adkinsrs]

@carlocolantuoni, how long ago was this command attempted? The error logs (going back to Jan 10) do not have anything related to the particular API call, so this was either run before then or it is a Javascript error. If it's a Javascript error, I cannot replicate as the dataset is private and clicking the link does not give me access to the dataset. My first inclination, though, is that maybe one of the Seurat variable names with a dot was not escaped in the javascript code.

[carlocolantuoni]

i did this all jan 22

i tried again just now and got no error ,but it didnt progress to the plot page this time, just eternally (waited 5 min or so) spinning when i hit the plot button

[adkinsrs]

Sounds to me that it is a Javascript issue.

Can you right click the page, and click Inspect at the bottom of the dropdown. This should bring up a developer console with various tabs. Next, click the "console" tab, and send a screenshot of the results. This should help me figure out where the javascript error is. If possible, if you see text in a red background, you can click the side arrow to the far left of the text to bring up a stacktrace.

[adkinsrs]

Alternative, what happens if you attempt to create the plot with none of the clusterbar groups selected? This may help me identify if I did not escape those group names.

[carlocolantuoni]

with no clusterbar groups seleceted im getting what i got in the 1st error report in this ticket:

[carlocolantuoni]

here's what i see when i "inspect" the page (sing original settings with clsterbar for this):

[carlocolantuoni]

and with some of those red text arrows expanded:

[carlocolantuoni]

any of that helpful? sorry got go for now, but will check back tomoro

[adkinsrs]

So I think you missed a step. After you bring up the Inspect window (which show the HTML as you screenshotted), there are tabs at the top of the pane. One of them is "console", and clicking that should bring up any potential Javascript errors. There should also be a sideways triangle to the left of the text in the red box that you can expand to bring up the stack trace information (to get to the right lines of code).

At least we eliminated escaping clusterbar group names as the potential culprit. If this is all too complicated, you could just give me this set of user credentials, so that I may be able to see the dataset and debug myself.

[carlocolantuoni]

sorry i missed that - this is all that comes up when i hit the console button:

[jorvis]

I've sent the credentials to Shaun but let's avoid posting account info on public tickets. :)

[carlocolantuoni]

aaaah - sorry!
thnx @jorvis
did not realize this was all public

[adkinsrs]

@carlocolantuoni

FIRST ATTEMPT
I think this may be the error (at least when I tried two random genes from a gene list this user saved).

Only one gene from the searched gene symbol was found in dataset. The heatmap option requires at least 2 genes to plot.

If ortholog mapping cannot find a map for a requested gene in the dataset then the gene will not be used. And all multigene plots require at least 2 genes, with heatmaps explicitly failing if only one gene mapped due to restrictions in the 3rd party function used to make the heatmap.

From the developer perspective, I need to maybe do a pre-mapping before "Submit" is allowed to be hit, as well as ensure this particular API error is communicated back instead of a generic one.

SECOND ATTEMPT
I used some more common genes (Pou4f3, Atoh1, Pou4f1), and the API job was killed due to running out of memory (256 Gb of RAM on the machine). Usually I see this happen with heatmaps that have clustering enabled, but this did not. I remember resolving a siimilar issue with heatmaps in the past due to the sheer number of metadata permutations, but I'm not sure if that was the reason here.

After filtering to just two main.lineaage categories, I saw the same error I saw in the first attempt. I switched to a dotplot and saw only Pou4f3 out of the 3 genes were present.

SUMMARY
Given how much involvement you have had with the dataset, I assume you know what genes are in it, so I think you just saw it run out of memory attempting to create the heatmap. For now I would suggest either filtering down the data (on the post-plot page) and hitting "Update", or create a dotplot instead which has no memory issues

As for me, I am going to create a JIRA ticket with the various suggested changes I noticed