search

lrgr / explosig

ExploSig: web-based interactive visualization of mutation signatures and somatic mutation data
https://explosig.lrgr.io
MIT License
1 stars 2 forks source link

Individual Sample Exposure Panel Doesn't Update After New Signatures Are Added #86

Open wir963 opened 5 years ago

wir963 commented 5 years ago

Example workflow. Use the data for reproducing the ERCC2 paper using the default signatures. Toggle on cosine similarity. Examine the single sample exposure panel for the TCGA-DK-A6AW-01A-11D-A30E-08, the sample with the most mutations (notice the cosine similarity between reconstructed mutation spectrum and true mutation spectrum is ~.4). Next, add Signature 10 and update. You should see the cosine similarity for TCGA-DK-A6AW-01A-11D-A30E-08 is now ~.97. Next, click on TCGA-DK-A6AW-01A-11D-A30E-08 and view in the single sample exposure panel. Some components, like the cosine similarity are correct. Other components like the exposure to signatures are not updated. Weirdly, when I add the mutation status of POLE, the mutation status of POLE is silent in the individual sample viewer but missense in the overview viewer

keller-mark commented 5 years ago

I think this is two separate issues:

  1. That sample (TCGA-DK-A6AW-01A-11D-A30E-08) has both POLE Silent and POLE Missense mutations according to the gene-alterations file. The same behavior occurs for sample (TCGA-ZF-A9RC-01A-11D-A38G-08) which shows missense in the individual-sample viewer and silent in the all-sample viewer. Sorting must be occurring differently such that it selects differently based on the plot.
screen shot 2019-02-12 at 5 20 48 pm screen shot 2019-02-12 at 5 23 33 pm

So we should decide what to do when a sample has two differing mutation categories for the same gene.

  1. The exposures to the signatures not updating should be an easy fix (I think it is caching the previous requests which I can change to be cleared upon updated selections)

Thanks for catching these!

wir963 commented 5 years ago

Thanks for looking into this @keller-mark! I made a new issue for the first part of this issue (see #87) and it sounds like you know what needs to be done for the second part.