search

iobio / cohort-gene

1 stars 0 forks source link

Crashing on TTN and DMD sized genes #1

Open stefinfection opened 6 years ago

stefinfection commented 6 years ago

This is currently only happening with large genes due to memory overflow on client side. A good solution to this is to cull the variant information being pulled back and optimize the data structures holding what we really need.

AlistairNWard commented 6 years ago

One good test is to look at the TTN gene. Very big and can often cause problems. I just ran TTN and crashed cohort and hub.

stefinfection commented 6 years ago

This is happening before the stream even finishes returning, making me believe too many variants are coming back. The solution to this will be filtering the stream. There are a few ways we can tackle this:

Tell the user there are too many variants to display in an unfiltered manner for the locus and

  1. display half of the gene

    • This will require giving them some ability to go to the next part of the gene (a slider at top?)
    • Caveats: half may still be too large of an area, would we repeat halving until an appropriate size is reached? Don't want to prompt the user twice...

  2. display high impact variants

    • Caveats: could still be too many variants in rare case; may not be what user wants
    • How will we modify further filtering to work when we're working with limiting the variants? Will have to make multiple annotation calls without changing the area which may appear clunky...
AlistairNWard commented 6 years ago

Rather than returning half the gene, which seems arbitrary, would it make sense to start with only returning variants within exonic regions?

Or can we send multiple requests to bring in the variants in stages? If d3 is drawing the variants as they are returned, would it even be noticeable if the variants were returned from multiple requests?