Performance on genes with targeted cancer therapies

[dhimmel]

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]

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

[brankaj]

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]

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

[KT12]

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]

@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]

Understood!

[KT12]

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

[dhimmel]

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

[KT12]

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]

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]

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