search

VALIS-software / valis-hpgv

High performance WebGL genome visualization
http://www.valis.bio
MIT License
14 stars 5 forks source link

Signal Track: Settings UI #17

Open haxiomic opened 5 years ago

haxiomic commented 5 years ago

Currently auto-scale is always on, would be nice to make that easily configurable by the user

Additionally it would be good to work out a consistent track display settings interface

hitz commented 5 years ago

Autoscaling is great to prevent "blank" tracks but actually can be really scientifically misleading; at low scales you are just looking at noise.

There should be a per-track option to set a scale for viewing; ideally this would be dynamic so if someone really wanted to see it autoscale or detect that no pixels were drawn and suggest that you turn on autoscaling then a UI button could change it. Anshul says that he can give us good default signals per data type/track type.

haxiomic commented 5 years ago

Hey @hitz, I think it's a good idea to have a UI option to control scaling. You can currently do it manually in the track configuration by setting autoScale and scale. For example to disable auto-scaling and set scale to 2x:

{
    name: 'Cerebellum, DNase',
    type: 'signal',
    path: "https://www.encodeproject.org/files/ENCFF833POA/@@download/ENCFF833POA.bigWig",
    autoScale: false,
    scale: 2,
}

(The values of the signal are not changed by the 2x scale, just the display – so the y-axis now ranges from 0 to 0.5 rather than 0 to 1)

Do you think a logarithmic display would be a good default for most signal data? Most data I see seems spiky and it's hard to capture the full range when displayed linearly

hitz commented 5 years ago

oh good to know. I've never seen a log scale. Some machine learners apply an sinh^-1 (inverse hyperbolic sin) to increase signal to noise ratio. I don't know what it looks like graphically, though.