Fix gene network graph discrepancy with old adage - Githubissues

greenelab / adage-frontend

The Adage web app, a tool to explore gene expression data and discover new insights from machine learning models

https://adage.greenelab.com

BSD 3-Clause "New" or "Revised" License

4 stars 3 forks source link

Fix gene network graph discrepancy with old adage #193

Closed vincerubinetti closed 4 years ago

vincerubinetti commented 4 years ago

modify gene graph construction and filtering code
write tests for gene graph construction and filtering code
dont clear search box when pressing enter to select
updates test dummy data to reflect small gene changes to database

vincerubinetti commented 4 years ago

The calculate enriched signatures test failed. The pValues changed slightly for some reason.

I did update the /dummy-data master genes list because I remember @dongbohu added some new standardName's, but I don't know why that would've affected the pvalues.

Any ideas @dongbohu? Did any of the gene ids change? Did the gene participation values change?

vincerubinetti commented 4 years ago

well i just updated the test to pass with the new values. you can still look here to see the differences: https://github.com/greenelab/adage-frontend/runs/766844937

dongbohu commented 4 years ago

The new adage server includes more genes, but the original genes have same IDs. Is the p-value depend solely on the dummy data?

vincerubinetti commented 4 years ago

The participation data of PA4638 is now included. On the old adage server, this gene didn’t exist in the database.

Your comment on slack. This probably explains the slight difference. One extra participating gene would be enough to slightly change the p value the way we're seeing. So I'm thinking it's okay.

dongbohu commented 4 years ago

To clarify, this gene's participation data was added a few weeks ago (before we migrated adage.greenelab.com).

vincerubinetti commented 4 years ago

To clarify, this gene's participation data was added a few weeks ago (before we migrated adage.greenelab.com).

Looks like the last PR for the frontend was merged ~22 days ago, so I guess that backend change was done slightly after that, which explains why none of my tests failed.