Linking Tracks for Autoscaling

[Ron]

Hi all,

First of all: thank you for all your hard work on IGV! It is much
appreciated.

Would it be possible to include a feature that would "link" two or
more tracks together such that the data range for all linked tracks is
the same at any given locus and scaled to the maximum and minimum
value of each track? For instance, if I were examining 3 different
ChIP-seq tracks for H3K4me3 and 3 different ChIP-seq tracks for
H3K36me3, I could link each set of 3 tracks together, autoscale them,
and scroll around the genome without having to reset my data range as
I pass over different genes, active vs. repressed regions, etc.

Thanks again,

Ron

[Jim Robinson]

You can't do that currently, but it is a good suggestion.  I've entered it in our issue tracker here

http://code.google.com/p/igv/issues/detail?id=46

thanks

Jim

[Matthew Anderson]

Hi Jim,

I believe a few of our uses have been wanting a similar feature for a
while now.
I.e. for sequencing data (bam) to be able to have an option to scale
all the coverage tracks the same.

Matthew

[James Robinson]

Hi Matthew, 

Yes I have several requests now similar to this one.    I'd like to solicit some ideas on the following questions.

How are the group(s) specified?  
How is the grouping shown, if at all, by that I mean how do you make clear for example A,B,and C are grouped, as are D and E?
How do you ungroup tracks?

We already have a concept of grouping using sample information,  so I don't want to introduce another ad-hoc concept, we'll all get confused eventually.   I welcome everyones ideas on this, if we can settle on a design the implementation should be straightforward.

Thanks

Jim

[Ron]

I don't know the policies of this group, so pardon me if this is
forward, but my suggestion is as follows:

Groups could be specified most easily by selecting >1 tracks, right
clicking and selecting something like "Link tracks and autoscale."
Then the tracks could automatically moved such that they are next to
each other in the viewer and in a separate panel. Finally, you could
display a "lock" icon in each track, perhaps with a line or "bracket"
that spans (vertically) the tracks that are linked. As long as the
tracks are linked, they should not be able to be moved to other
panels, except perhaps as group.

It's possible that some people will want to link tracks but not want
them displayed next to each other... I'm not sure if this can be
realistically done, especially if the user links multiple sets of
tracks.

Sorry again if that is forward.

Best,

Ron

On Jul 27, 8:26 am, James Robinson <jrobi...@broadinstitute.org>
wrote:

[James Robinson]

Ron, thanks for the suggestions,  this is exactly the kind of input I'm looking for and what this forum is for.

There is already a "grouping" concept that moves tracks next to each other,  with some space between groups, close to what you suggest.  The only thing missing is the linked auto-scale.   I'm assuming however that people will not neccessarily want their tracks moved, as you suggest,  and it will be difficult to keep track which tracks are linked to which.    Perhaps the requestors of this feature will chime in.

thx

Jim

[David Wood]

Hi Everyone yes I've been looking for a feature like this in IGV for
some time. It would especially be useful with RNA-seq data, when
viewed as either coverage tracks, or TDF files (barcharts or line
plots). I have many different tracks and data types for one sample,
and would only want certain tracks 'group autoscaled', but would
prefer that those tracks were not automatically clustered together in
the view (eg, could remain embedded in a 'sample' group, for
instance).

I like the idea of using the existing sample attributes framework for
defining tracks to join and 'group autoscale'. Perhaps in the View-
>Preferences->Charts panel, there can be an option below 'Continuous
Autoscale' called 'Group Autoscale' which contains a list of all
sample attributes, and provides for the selection of one or more to
'group autoscale'? This way, grouping is controlled through existing
mechanisms, is independent of the track order and view layout, and the
option is located next to other scaling options in the preferences
panel?

Cheers,
David

On Jul 28, 5:53 am, James Robinson <jrobi...@broadinstitute.org>
wrote:

[David]

Hi Everyone yes I've been looking for a feature like this in IGV for
some time. It would especially be useful with RNA-seq data, when
viewed as either coverage tracks, or TDF files (barcharts or line

plots). I have many different tracks for one sample, and would only

want certain tracks 'group autoscaled', but would prefer that those
tracks were not automatically clustered together in the view (eg,
could remain embedded in a 'sample' group, for instance).

I like the idea of using the existing sample attributes framework for
defining tracks to join and 'group autoscale'. Perhaps in the View-
>Preferences->Charts panel, there can be an option below 'Continuous
Autoscale' called 'Group Autoscale' which contains a list of all
sample attributes, and provides for the selection of one or more to
'group autoscale'? This way, grouping is controlled through existing
mechanisms, is independent of the track order and view layout, and the
option is located next to other scaling options in the preferences
panel?

Cheers,
David

On Jul 28, 5:53 am, James Robinson <jrobi...@broadinstitute.org>
wrote:

[Jim Robinson]

OK,  I will bump this up the prioritized feature list and try to get to it in August.   Thanks for the discussion. 

[Cook, Malcolm]

Perhaps some food for thought - An visualization in Gbrowse of multiple coverage grouped vectors: http://flybase.org/static_pages/feature/previous/articles/2011_06/new_GBrowse_tracks.html

[Malcolm Cook]

Hi....

I too find that linking combined with auto-scaling would be a great addition to IGV.

For my uses, I would be happy if it were only possible to do this with tracks that have been 'grouped' together, either manually or via the grouping mechanism.

However, we need more/better ways of assigning new attributes upon which samples can be grouped.

The only way now requires loading an attribute file.  If in addition we new attributes simply within IGV, I think you would meet most users needs.

Three ways of assigning new attibutes come to mind:

          1) use the 'basename' or 'extension' or 'directory' of the  DATA FILE 

          2) extract a portion of the NAME or DATAFILE using a regular expression with capture

         3) provide an interactive dialog

~ Malcolm

[Stu]

Our group would love to have this feature implemented to help look at stranded transcriptome data sets.  Has there been any progress, or has there been any work-arounds or patches that have been written by outside developers?

I like the idea suggested by Ron above where a "lock" icon is displayed if the track is grouped.   However, instead of moving the tracks together to distinguish the group, one could draw the "lock" icon using a unique color for the group ... 

thx,

-s.

[Malcolm Cook]

Jim,

Here is a suggestion.   Implement the following:

* new track function: "autoscale coordinated" which auto-scales selected tracks to their groupwise maximum.

This, in combination with existing (a) track grouping and (b) 'shift click' to selects all tracks in a group, gives a two-click recipe for group-wise scaling (albeit one group at a time).

Finally, if you implement multi-level grouping (i.e. just like you have multi-level sorting), I think MANY of the use-cases discussed in this thread would be satisfied.

IGV for the win,

Malcolm

On Wednesday, July 27, 2011 9:08:03 PM UTC-5, Jim Robinson wrote:

[Malcolm Cook]

Hi Jim, or the next guy.... I've just come back to IGV after a fairly long respite and still don't see any way to have the scaling of tracks be coordinated, possibly via some grouping mechanism as we're discussing here.  Now, I know I have this work around: when I link to a new ROI, I can hand select the tracks I want coordinately scaled, choose 'set data ranges' and take advantage of the fact that IGV pre-populates this dialog with the group max.   This is fine, but, tiresome.  Any hopes for a glimmer of improvement here?

[Jim Robinson]

Hope springs eternal...     I'll look at this again.

--

---
You received this message because you are subscribed to the Google Groups "igv-help" group.
To unsubscribe from this group and stop receiving emails from it, send an email to igv-help+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/igv-help/f2263440-f642-4b0c-8020-80c6bd49fb31%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Malcolm Cook]

Jim,

Indeed hope does spring eternal!

I found that for my particular use case creating a .gct file where the values were row normalized to sum to 100 fit the bill perfectly.  It allowed me to visualize each gene's relative expression  across its length in many samples at the same time.

IGV for the win,

Malcolm

~ Malcolm Cook
h: 913.642.5257
m: 913.226.1542
w: 816.926.4449

--

---
You received this message because you are subscribed to a topic in the Google Groups "igv-help" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/igv-help/O12r3RAKyck/unsubscribe.
To unsubscribe from this group and all its topics, send an email to igv-help+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/igv-help/558C782B.60603%40broadinstitute.org.

[Malcolm Cook]

Let me google that for you!

Is it woohoo yet!

[Jim Robinson]

Hi Malcolm,  its nearly woohoo.   Still to do are the mechanics for specifying the grouping.    If we use the current "grouping" mechanism you would have to physically group tracks together on the screen to autoscale.     

BTW,  if you have not done so could you send us a short "email of support"  for our grant,  and encourage friends and colleagues to do the same?  See the sticky notice on the igv-help group page.   These are very important.  

Thanks

Jim

[Malcolm Cook]

I note with glee and woohoo the arrival of group autoscale to IGV.

Alas I see no documentation for it at http://software.broadinstitute.org/software/igv/book/export/html/6 

Having an updated documentation would help me understand your intentions for group autoscale.

Currently  I am unsure as to whether the following observations are by design or not...

• There is no visual indicator as to which tracks are grouped
  
• I am unsure whether one set of tracks can be grouped together and another set of tracks can be grouped together (i.e. can I have two groups each subject to own group autoscale?)
  
• selecting "Group Autoscale" unselects "Autoscale" but it does not have its own check box.  Should it?

Thanks for travelling down this long road...

~Malcolm

[Helga Thorvaldsdottir]

Hi Malcolm,

When you've selected multiple tracks and clicked on "Group Autoscale" in the menu, you should see a new column appear in the attribute display panel (between the track names and the tracks). You can indeed have multiple groups, and each one will have a different color in the attribute display, so you can see which tracks are together in a group. In the screenshot below, there are 2 autoscale groups of 2 tracks each (green and brown in the attribute panel). Note that you can also select all the tracks in a group by clicking in the attribute panel.

Helga

[Malcolm Cook]

Hi,

I see it now.  Thanks.  Is this documented somewhere?  

FWIW: The label "AUTOSCALE_GROUP" is ~50% clipped on my display for some reason, making it hard to notice next to "NAME DATATYPE DATAFILE".

Also, the UI expression of the logical dependencies between "group autoscale" and "autoscale" are somewhat confusing to me.

For instance, it appears that, once group autoscale is enable for a set of tracks, it is not possible to turn off group autoscale without turning on individual autoscale.  

Thus if I want to entirely undo scaling for a subset of group auto-scaled tracks, I must first turn on individual autoscaling, and then turn it off.  

Also, I note if I create multiple autoscale groups they are not individually color-coded but all appear as dark blue, though they do scale seperately, and do display different group on hover.  I also see that clicking on a group selects all tracks in that group.  NICE!  Except for the absence of color.  (Maybe they are all just shades of dark blue, but they look the same to me).

I think this logic could be better re-inforced by the UI with the following change:

 * rename the GROUP AUTOSCALE track to simply be AUTOSCALE and have it always present

 * use WHITE which for tracks which dont autoscale

 * use BLACK for tracks which autoscale but are not in an autoscale group 

 * use COLORS (widely separated) to code autoscale group

But I do see that your autoscale groups are differently colored.

Hmmm - what's the difference?

Thanks for the education and the nice feature.  

~ Malcolm

~ Malcolm Cook
h: 913.642.5257
m: 913.226.1542
w: 816.926.4449

--

---

You received this message because you are subscribed to a topic in the Google Groups "igv-help" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/igv-help/O12r3RAKyck/unsubscribe.

To unsubscribe from this group and all its topics, send an email to igv-help+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/igv-help/bdc38867-171f-4318-902e-9e7f460ee3de%40googlegroups.com.

[Helga Thorvaldsdottir]

Malcolm,

Looks like you may be running an older version of IGV. The issue with the coloring of the group autoscale attributes was fixed a while ago.

Thanks for the UI improvement suggestions. I created a feature request in our GitHub.

And unfortunately these details are not yet in the User Guide. There was just a brief announcement in the release notes when group autoscale was released. 

Helga

[Malcolm Cook]

Indeed, updating fixes this, thanks!

And thanks for taking my suggestions....

[Malcolm Cook]

Is there now perhaps a issue to document group autoscale?

On Tuesday, May 2, 2017 at 10:55:10 AM UTC-5, Helga Thorvaldsdottir wrote: