It was not hard to find an example that highlights the limits of the linear scaling.
ALK in Nb has~25x greater expression vs BRAF and this expands the y-axis so that the GTEX sample cohorts’ boxes are compressed and don’t show a difference between sample types.
Following are potential solutions:
Keep linear scale. Discard extreme outliers from the plot, by setting y-axis upper limit to 10 * median, only if there is any extreme outlier, say max / median > 100.
Use log scaled y-axis if there is any extreme outlier, say max / median > 100.
Add an API query parameter for linear or log scaled y-axis. PedOT front-end delivers both, and users can switch between log and linear.
This issue will only be worked on after #21 is resolved.
This issue is directly related to #22 .
cc @jharenza @komalsrathi @taylordm @chinwallaa @jonkiky
As pointed out by @dunnpa in a slack message and #22,
Following are potential solutions:
Keep linear scale. Discard extreme outliers from the plot, by setting y-axis upper limit to 10 * median, only if there is any extreme outlier, say max / median > 100.This issue will only be worked on after #21 is resolved.
This issue is directly related to #22 .
cc @jharenza @komalsrathi @taylordm @chinwallaa @jonkiky