Handle genes that are longer than max-LZ-region

[pjvandehaar]

gls = [l.strip().split() for l in open('genes.bed')]
gls
gls2 = [(int(g[2]) - int(g[1]), g[0], g[3], g[4]) for g in gls]
len([g for g in gls2 if g[0] > 500e3])
# 311
sorted(gls2)[-1]
# (2304996, '7', 'CNTNAP2', 'ENSG00000174469')

Indeed, http://pheweb.sph.umich.edu/gene/CNTNAP2 is broken. How do we fix this?

Option 1: zoom in on part of gene with significance (Nah)

For that, we either must:

• record where the significance is
• go back and look it up when the page is requested

Option 2: do binning/subsetting for LZ

• Option 2a: subset in LZ rendering. (but hold onto all data)

• Option 2b: subset in api.
  LocusZoom re-requests every time the region changes, even when you zoom in, so this should work great.
  Heap the top 1000 variants for all LZ requests. Remove the region limit. Include {num_dropped_variants: 3500000, best_dropped_pval: 2.3e-3} in the response and display that somewhere. (Maybe dump it in Layout.state and put that in dashboard?