Re: [Gmod-ajax] quantitative ("wiggle") tracks do not work

[Mitch S. <mit...@be...>]

Brenton Graveley wrote:
> This would be good. But I still think the best option is a dynamic 
> scaling based on the max/min in the viewing window even though I 
> realize it is difficult to implement. But perhaps having 3 different 
> scales of images such as 1-100, 1-1000 and 1-10000 would suffice for 
> most cases?

I hear you; I've been thinking about it off and on since you first 
proposed this, but I still don't entirely get it.  So I'd like to go 
into some detail, because I think the answers to these questions will 
help me understand the use case a bit better:

Suppose there is dynamic scaling.  Say you're looking at wiggle data in 
one region where the range of values is 1-100 (plotted, say, 100 pixels 
high), and variation between (say) 80 and 90 is relevant and interesting 
to you.  Now suppose you scroll a teeny bit to the right, and suddenly a 
data value of 10,000 becomes visible.  Then (because of the dynamic 
scaling) the scale changes so that the range of 1-10,000 is plotted on 
the same 100 pixel high region, and now your 80-90 variation is a .1 
pixel difference and is invisible.  Isn't that a problem?

If you're not actually interested in the 80-90 range, then you can plot 
on a 1-10,000 linear scale and be happy.  If you really are interested 
in both 80-90 and also in 8,000-9,000 then you can plot on a log scale 
and be happy.

Is there a specific situation where those two options aren't enough?  It 
seems to me that the point of plotting quantitative data in a genome 
browser is to compare values in one area to values in another area, so 
shouldn't you be able to plot your data all on the same scale?

If you want to call extra attention to really high values we could 
always color them differently or something.  Also, I think it makes 
sense to give the user the option to "zoom" in the y-direction in a 
limited sense.  But given how hard dynamic scaling would be to 
implement, I think it'll be a while before I get to it.  But again, it's 
open source--if someone wants to tackle it I'd be happy to talk about 
how to do it.

> Another feature that would be nice is a scale bar or number to easily 
> show the amount of sequence shown in the current window?

I was just looking at UCSC and I noticed they have both, and GBrowse 
shows the number of bases.  I added a ticket for this.

Regards,
Mitch