Add in step after protein processing to do imputation

PNNL-CompBio / decomprolute

A suite of scientific workflows to assess metrics to compare efficacy of protein-based tumor deconvolution algorithms.

MIT License

15 stars 4 forks source link

Add in step after protein processing to do imputation #17

Closed sgosline closed 3 years ago

sgosline commented 3 years ago

Currently we are pulling just the normalized proteomics from CPTAC. We need to compare the performance of the algorithms on imputed data using https://github.com/WangLab-MSSM/DreamAI

This requires the DreamAI team assemble a docker image that we can then call using CWL. Can someone from the DreamAI team join the call?

sgosline commented 3 years ago

Anna will create Docker image for DreamAI.

annapamma commented 3 years ago

Notes from 12/8 meeting:

DreamAI can impute isoform data -- we may want to keep things as isoforms for as much of the process as possible to ultimately support tumor deconvolution for isoform data
We can add a fast option to the DreamAI script. This option will skip missForest imputation, which is the longest-running process

sgosline commented 3 years ago

I believe this was closed by #39. @LifeWorks can you confirm?

LifeWorks commented 3 years ago

Error in DreamAI(data, k = 10, maxiter_MF = 10, ntree = 100, maxnodes = NULL, : Some Rows Are Missing All Values !!

When I try to use the docker image to process the proteomics data, I find this error. @annapamma

It seems that DreamAI will check the na values and will stop if there are rows has all na values?

annapamma commented 3 years ago

Hey @LifeWorks -- can you send me the data you're testing this on?

If it's just the raw CPTAC data from the Python package, I can just go ahead and test that myself on each of the tumor types.

LifeWorks commented 3 years ago

@annapamma

I just downloaded the proteomics data using the scripts.

python protData/protDataSetsCLI.py --cancerType ccrcc

annapamma commented 3 years ago

Fixed with https://github.com/sgosline/proteomicsTumorDeconv/pull/51