search

cognoma / machine-learning

Machine learning for Project Cognoma
Other
32 stars 47 forks source link

Performance on genes with targeted cancer therapies #72

Open dhimmel opened 7 years ago

dhimmel commented 7 years ago

It seems that some of the most interesting genes to predict using expression will be genes that have targeted therapies. These genes are ideal, because we may be able identify unmutated samples that would still respond to the therapy based on their expression.

Here's a list of targeted cancer drugs and thier genes. Tagging @brankaj.

KT12 commented 7 years ago

Would it be a good idea to add a new folder attempting to predict the genes on that list?

brankaj commented 7 years ago

Hey @KT12 , I believe we can still upload the files to this folder: https://github.com/cognoma/machine-learning/tree/master/explore. Maybe we can name files based on the different subset of genes.

KT12 commented 7 years ago

@brankaj , sounds good. Sorry I didn't realize what the folder was for.

KT12 commented 7 years ago

Perhaps naming convention could be:

Classifier_Gene_Github-handle.ipynb where Gene = column head from mutation-matrix.tsv.bz2?

@dhimmel would you like to weigh in?

dhimmel commented 7 years ago

@KT12: I'm not sure you need a new notebook for every mutation. It could be a good time to start migrating to the cognoml package for machine learning, in which case you could perhaps do multiple models in a single notebook.

KT12 commented 7 years ago

Understood!

KT12 commented 7 years ago

I experienced some issues doing pip install --editable . for cognoml, but I guess we can discuss this at the meetup tonight.

dhimmel commented 7 years ago

@KT12 can you post the error as a new issue on https://github.com/cognoma/cognoml/issues

KT12 commented 7 years ago

I would but it might be my own fault. I git clone'd cognoml to a new directory and then did pip install --editable . By process of elimination, I think it may have messed up other package dependencies (scipy and nltk), but I'm not certain. So I'd like to reattempt installing in the presence of experts before raising a new issue.

dhimmel commented 7 years ago

So I'd like to reattempt installing in the presence of experts before raising a new issue.

No problem. In general I think it's more important to keep discussion of an open issue on topic. Opening a new issue isn't a big deal as it's easy to close it in case of use error. And if there's actually a bug (which there often is), we want a record. We'll look into this tonight.

KT12 commented 7 years ago

Re-installed in the cognoma-machine-learning environment and it hasn't interfered. So chalk my problem up to user error.