As a continuation of #48, we ran the same experiments for more genes, and added random feature selection (of the same number of features) as a baseline.
General results are similar to what we saw before - PTEN is an example of a gene where our feature selection method (median_f_test) works well for non-carcinoma cancer types:
Some other genes we tried (RB1, BRAF) don't seem to work as well, though.
I also added a script at 02_cancer_type_classification/scripts/run_fs_for_genes.sh to run 02_cancer_type_classification/plot_univariate_fs_results.ipynb for all the genes we looked at and save the resulting plots (I didn't track the plots here, just locally).
As a continuation of #48, we ran the same experiments for more genes, and added random feature selection (of the same number of features) as a baseline.
General results are similar to what we saw before - PTEN is an example of a gene where our feature selection method (
median_f_test
) works well for non-carcinoma cancer types:Some other genes we tried (RB1, BRAF) don't seem to work as well, though.
I also added a script at
02_cancer_type_classification/scripts/run_fs_for_genes.sh
to run02_cancer_type_classification/plot_univariate_fs_results.ipynb
for all the genes we looked at and save the resulting plots (I didn't track the plots here, just locally).