Some proposals from users

[kokitsuyuzaki]

Some proposals from users

[kokitsuyuzaki]

Congratulations on accepting your paper!
https://joss.theoj.org/papers/10.21105/joss.05603

[jkanche]

faceting is on our list of things to do (#199) for the next release.

We are trying to clear up our plate and hopefully start on the new version soon!

[jkanche]

Thanks for promoting kana! I really appreciate the feedback and we'll try to incorporate these in the upcoming releases.

We are thinking of some of these, especially faceting by annotation for the next release, so stay tuned... #199

[kokitsuyuzaki]

This is my proposal.

Version changes may cause kana suddenly stop working as follows.
#246

Could you please keep the previously versions of kana somewhere?
This is because I had trouble identifying whether the error is on the tool side or the data side.

Although I know you are keeping the version 2 (https://www.jkanche.com/kana/), I want to compare the behavior between different versions of kana.
For this purpose, it would be nice to be able to track down to a finer version like

https://www.kanaverse.org/kana/v3.0.1
...
https://www.kanaverse.org/kana/v3.0.14
https://www.kanaverse.org/kana/v3.0.15
https://www.kanaverse.org/kana/v3.0.16
https://www.kanaverse.org/kana/v3.0.17
https://www.kanaverse.org/kana/v3.0.18
https://www.kanaverse.org/kana/latest
https://www.kanaverse.org/kana/devel
...etc

[jkanche]

This is my proposal.

Version changes may cause kana suddenly stop working as follows. #246

Could you please keep the previously versions of kana somewhere? This is because I had trouble identifying whether the error is on the tool side or the data side.

Although I know you are keeping the version 2 (https://www.jkanche.com/kana/), I want to compare the behavior between different versions of kana. For this purpose, it would be nice to be able to track down to a finer version like

https://www.kanaverse.org/kana/v3.0.1 ... https://www.kanaverse.org/kana/v3.0.14 https://www.kanaverse.org/kana/v3.0.15 https://www.kanaverse.org/kana/v3.0.16 https://www.kanaverse.org/kana/v3.0.17 https://www.kanaverse.org/kana/v3.0.18 https://www.kanaverse.org/kana/latest https://www.kanaverse.org/kana/devel ...etc

This is a good idea, We'll update the ci/cd pipeline to keep all version of the app rather than only the latest.

[kokitsuyuzaki]

Looking forward to the next release!
There were two more things I forgot to mention.

• Name of each analysis mode

I see that there are four modes of analysis listed below:

1. I want to analyze a new dataset
2. I want to explore existing analysis results
3. I just want to try it out
4. I want to go back to the old version!

It would be convenient if each mode has some name, for example, I know the second one is called "explore mode", so I would like appropriate names for the remaining three.

• Pre-loading dataset

In the third mode "I just want to try it out", the input data has been already loaded.
I don't know the details but if it can also configured to load user's original data, it is really helpful.
The current kana is convenient enough but if this can be achieved, the results of data analysis can be viewed by simply giving the URL to my collaborators, not providing both the data and the URL https://www.kanaverse.org/kana/

Thank you for your consideration!

[LTLA]

Of course, I know kana is Javascript-based, so you can't use schex directly, but it would be great if you could sort cells so that the more expressed they are, the more they come to the front of the screen.

iSEE/iSEE#627 may be worth reading here, and specifically my response.

Technically, we should (deterministically) shuffle the cells before plotting them, and this would be the most statistically correct approach. I suppose we could use "sort by" as an option, but not the default, to avoid misleading conclusions.

[kokitsuyuzaki]

Thanks, @LTLA
It seems not a trivial problem.

As @federicomarini and @LTLA mentioned, the cell sorting ignores the distribution of gene expression and overestimates highly expressed genes, and can be a trade-off between false positive and false negative.

In any case, it is difficult to visualize cells in two dimensions when the number of cells is large, so it may be necessary to summarize them by region like hexagon or meta cells combined by neighboring cells, etc.

[kokitsuyuzaki]

• When the number of cells is large, the selected cell hides behind other cells

My collaborator said that his situation is more complicated; the input data has 10 batches (e.g., batch1 to batch10) and cells in batch1 always hide behind other batch's cells but cells in batch10 always come to the front of screen, which implies that cells are visualized according to the order of batch IDs.

Or, this might be due to the order of columns (e.g, col1, col2, ...) in input matrix, because this data is created by concatenation of 10 data matrices.

[kokitsuyuzaki]

The following requests were also made by my collaborator.

• Cell selection based on multiple gene expression

It would be nice to be able to sum the expression of multiple genes, or take the product set of cells selected by CUSTOM SELECTIONS.
This functionality contributes to stratify cells based on multiple marker genes.

Also, he found a behavior that's probably a bug.

• Switching the versus arrow does not change the gene list

After selecting two regions in Selection,
two cell groups are automatically labeled such as cs1 and cs2.
When he compared two groups by "versus" button,
differentiated genes between these groups were listed,
but pressing the ⇆ button did not cause the re-sorting of the genes.

[maggieeiggam]

Thanks, @LTLA It seems not a trivial problem.

As @federicomarini and @LTLA mentioned, the cell sorting ignores the distribution of gene expression and overestimates highly expressed genes, and can be a trade-off between false positive and false negative.

In any case, it is difficult to visualize cells in two dimensions when the number of cells is large, so it may be necessary to summarize them by region like hexagon or meta cells combined by neighboring cells, etc.

+1 for a smoothed summary

Other simple options could be to decrease the point sizes, and add faceting.