Significant threshold setting on Localzoom

117 views
Skip to first unread message

Lilian

unread,
Jul 15, 2021, 11:04:06 AM7/15/21
to locuszoom
Hi everyone,

I am currently concerned by the automatic setting of the significant threshold in Localzoom. I would like to be able to change it, or to discard it if it's not possible. This task seems simple, but i haven't been able to find where and how to arrange that, is it that i overlooked something or is it really impossible on Localzoom ?

Thank you for your attention,
Lilian GOEPP

Andy Boughton

unread,
Jul 15, 2021, 4:08:57 PM7/15/21
to locu...@googlegroups.com
Thanks for the question!

We'd like to build user-friendly features for controlling the plot, but at present there is no UI to control the default plotting options. 

That said, it can be done if you are familiar with your web browser's "JavaScript console", via the following magical incantation that directly calls custom code on the page. This changes the default plotting options for the association line of GWAS significance:

lz_plot_1.layout.panels[0].data_layers[0].offset = 1.234;  // Default: is 7.301, also known as -log10(.05/1e6). Each association plot is a separate panel. If you are showing two plots, you may change panels[0], panels[1], etc
lz_plot_1.applyState(); // Re-render the plot, which causes layout changes to be applied


Obviously, this is not very intuitive. One thing that people can do to help us improve our UI is to get feedback on the kinds of data or defaults that would be helpful: LocusZoom has an infinite combination of options for color, size, shape, and what kind of data to show. But people would probably not like having to wade through infinite buttons. Knowing which settings are most important to change would help us focus on the most useful and user-friendly options first!

The line of significance is definitely on the short list: https://github.com/statgen/locuszoom/issues/252 

-Andy Boughton
abo...@umich.edu

Senior Applications Programmer/Analyst
Center for Statistical Genetics
University of Michigan

--
You received this message because you are subscribed to the Google Groups "locuszoom" group.
To unsubscribe from this group and stop receiving emails from it, send an email to locuszoom+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/locuszoom/b8a4150f-a02a-45cd-8f98-d8bb45b1de8fn%40googlegroups.com.

Lilian

unread,
Jul 16, 2021, 12:38:08 PM7/16/21
to locuszoom
Thank you for this answer.

It could be great to add this "Javascript console" trick to the instructions of localzoom, it does indeed seems powerful. I'm still having a little issue though : as the plot had to include a high (in -log10) significant threshold, the whole graph is streched to include it, which leads to the data being smashed toward the abscissa. Changing the significant threshold does not change this fact unfortunately. Therefore, can i still ask you how to change the disposition of the graph with respect to the log10 ordinate ? Like setting the minimum and maximum ordinate.

I added an artifical example so that my issue is more clear :

locuszoom_example_image.png

Here, the graph is clearly displayed as if the significant threshold was still 7.3, hence the stretching of the graph.

Thank you again for your help,
Lilian GOEPP

Andy Boughton

unread,
Jul 16, 2021, 12:57:24 PM7/16/21
to locu...@googlegroups.com
Another obscure feature (wheee!):

If you click-and-drag one of the y axis value/tick labels, you can zoom in along the y scale. (you'll see your mouse pointer change when you hover in the right place) The default range is -log10p = 0...10. I agree that small values can be hard to see on this range... this default value was chosen on the theory that very small values are usually below the line of GWAS significance, and thus, having a minimum range kept people from accidentally focusing too much effort on "peaks" that had limited scientific meaning.

Again, of course, this reflects assumptions that most people are visualizing GWAS data, with defaults that are based on domain knowledge around that specific problem. Other data types may well have lower line of significance, in which case zooming in would be very appropriate.

This really reflects the fact that LZ.js was written for data sharing portals, where the website developer knew in advance what would be shown and could customize once to serve many users. Sites like my.locuszoom.org are intended to be more reusable, so we follow discussions about new kinds of data / per-user customization with interest. :) (we touch on the difference between command line and web tools a bit in the supplementary section of the recent LZ.js paper: "These advantages encourage collaboration and reduce artificial barriers to entry from command line syntax, but tend to emphasize solving the most common problems at the expense of customization in the most advanced use cases")

-Andy Boughton
abo...@umich.edu

Senior Applications Programmer/Analyst
Center for Statistical Genetics
University of Michigan



On Jul 16, 2021, at 12:38 PM, 'Lilian' via locuszoom <locu...@googlegroups.com> wrote:

Thank you for this answer. 

It could be great to add this "Javascript console" trick to the instructions of localzoom, it does indeed seems powerful. I'm still having a little issue though : as the plot had to include a high (in -log10) significant threshold, the whole graph is streched to include it, which leads to the data being smashed toward the abscissa. Changing the significant threshold does not change this fact unfortunately. Therefore, can i still ask you how to change the disposition of the graph with respect to the log10 ordinate ? Like setting the minimum and maximum ordinate.

I added an artifical example so that my issue is more clear :

To view this discussion on the web visit https://groups.google.com/d/msgid/locuszoom/4f8bcf82-cfb2-48d4-9d26-358aafe67377n%40googlegroups.com.
<locuszoom_example_image.png>

Lilian

unread,
Jul 16, 2021, 1:30:00 PM7/16/21
to locuszoom
God i feel dumb ! Anyway, thank you a lot for your availability !
Reply all
Reply to author
Forward
0 new messages