Skip to content

Small improvements - #11

Open
anderdnavarro wants to merge 5 commits into
LincolnSteinLab:mainfrom
anderdnavarro:main
Open

anderdnavarro wants to merge 5 commits into
LincolnSteinLab:mainfrom
anderdnavarro:main

Conversation

@anderdnavarro

Copy link
Copy Markdown

I have changed three things:

  1. I have added the option to allow the user to change the default predictions_DeepTumour.json output name with the --outName tag
  2. LiftOver chain file was prepared for chr1 prefix style chromosomes only. I have added a new file without the chr prefix.
  3. In the Readme, the Local Environment section specifies Python 3.10, but we use Python 3.9 and there are some packages that are not supported for Python 3.10. I have specified Python 3.9 because it's also the one used for the Docker image.

@lstein

lstein commented Oct 1, 2026 via email

Copy link
Copy Markdown

@anderdnavarro

Copy link
Copy Markdown
Author

With the last two commits, I have:

  1. Added debugging files to build a Dev Container in VS Code and use its default debugpy engine for faster and easier debugging.
  2. Fixed an error while filtering the VCF that caused DeepTumour v3.0.5 to always predict "CNS-PiloAstro".

In more detail, the last two DeepTumour updates used an extra filtering step to keep only "PASS" mutations. However, the allel.vcf_to_dataframe() function we use to read the VCF considers the field FILTER == '.' as FAIL. Having FILTER == '.' is common in many VCFs, and OncoGAN produces VCFs in this format. Therefore, I have added a step to update the FILTER field from . to PASS before using the allel.vcf_to_dataframe() function, avoiding the removal of all mutations in these cases.

@anderdnavarro

Copy link
Copy Markdown
Author

I have checked that this version works and have nothing else to add. If everybody agrees, I think this pull request is ready to be merged.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants