plot_metric = 'auroc'
images_dir = Path(cfg.images_dirs['multimodal'], 'auroc')
else:
plot_metric = 'aupr'
images_dir = Path(cfg.images_dirs['multimodal'])
# ## Results with compressed features (figures in main paper)
#
# We'll also look at results with raw features later, those figures go in the supplement.
#
# ### Compare raw results
# In[3]:
# load raw data
results_df = au.load_stratified_prediction_results(results_dir, 'gene')
# drop TET2 for now
results_df = results_df[~(results_df.identifier == 'TET2')].copy()
# make sure that we have data for all data types and for two replicates (random seeds)
print(results_df.shape)
print(results_df.seed.unique())
print(results_df.identifier.unique())
print(results_df.training_data.unique())
results_df.head()
# In[4]:
# each subplot will show results for one gene
sns.set({'figure.figsize': (24, 12)})
import mpmp.utilities.plot_utilities as plu
# In[2]:
# set results directory
old_results_dir = Path(cfg.results_dirs['mutation'], 'methylation_results',
'gene').resolve()
new_results_dir = Path(cfg.results_dirs['mutation'],
'methylation_results_shuffle_cancer_type',
'gene').resolve()
# In[3]:
# load raw data
old_results_df = au.load_stratified_prediction_results(old_results_dir, 'gene')
new_results_df = au.load_stratified_prediction_results(new_results_dir, 'gene')
# here we want to use compressed data for methylation datasets (27k and 450k)
# the results in 02_classify_compressed/compressed_vs_raw_results.ipynb show that
# performance is equal or slightly better for PCA compressed methylation data,
# and it's much easier/faster to fit models on
old_results_df = old_results_df[old_results_df.training_data.isin(
['expression'])].copy()
new_results_df = new_results_df[new_results_df.training_data.isin(
['expression'])].copy()
old_compressed_results_df = au.load_compressed_prediction_results(
old_results_dir, 'gene', old_filenames=True)
# load compressed data for me_27k and me_450k
old_compressed_results_df = old_compressed_results_df[
# if True, save figures to ./images directory
SAVE_FIGS = True
# if True, plot AUROC instead of AUPR
PLOT_AUROC = False
if PLOT_AUROC:
plot_metric = 'auroc'
images_dir = Path(cfg.images_dirs['mutation'], 'auroc')
else:
plot_metric = 'aupr'
images_dir = Path(cfg.images_dirs['mutation'])
# In[3]:
# load raw data
vogelstein_df = au.load_stratified_prediction_results(vogelstein_results_dir,
'gene')
vogelstein_df = vogelstein_df[vogelstein_df.training_data.isin(['expression'])]
vogelstein_df['gene_set'] = 'vogelstein'
# make sure that we're correctly pointing to raw data for non-methylation data types
# and that we have data for two replicates (two random seeds)
print(vogelstein_df.shape)
print(vogelstein_df.seed.unique())
print(vogelstein_df.training_data.unique())
vogelstein_df.head()
# In[4]:
# load raw data
top_50_df = au.load_stratified_prediction_results(top_50_results_dir, 'gene')
top_50_df = top_50_df[top_50_df.training_data.isin(['expression'])]
# set results directory # this is a mess, TODO move results into same location results_dir1 = Path(cfg.results_dirs['mutation'], 'bmiq_results', 'gene').resolve() results_dir2 = Path(cfg.results_dirs['mutation'], 'bmiq_results_2', 'gene').resolve() results_dir3 = Path(cfg.results_dirs['mutation'], 'bmiq_results_me_control', 'gene').resolve() # set significance cutoff after FDR correction SIG_ALPHA = 0.001 # In[3]: # load raw data results_df1 = au.load_stratified_prediction_results(results_dir1, 'gene') results_df1.loc[results_df1.training_data == 'me_27k', 'training_data'] = 'me_27k_corrected' results_df2 = au.load_stratified_prediction_results(results_dir2, 'gene') results_df2.loc[results_df2.training_data == 'me_27k', 'training_data'] = 'me_27k_corrected' results_df3 = au.load_stratified_prediction_results(results_dir3, 'gene') results_df = pd.concat((results_df1, results_df2, results_df3)) print(results_df.shape) print(results_df.seed.unique()) print(results_df.training_data.unique()) results_df.head() # In[4]:
from adjustText import adjust_text
import mpmp.config as cfg
import mpmp.utilities.analysis_utilities as au
# In[2]:
# set results directory
results_dir = Path(cfg.results_dirs['cancer_type'], 'cancer_type').resolve()
# set significance cutoff after FDR correction
SIG_ALPHA = 0.001
# In[3]:
# load raw data
results_df = au.load_stratified_prediction_results(results_dir, 'cancer_type')
print(results_df.shape)
results_df.head()
# In[4]:
# set this variable to filter plot to certain cancer types
# if not included, plot all 33 of them
# FILTER_CANCER_TYPES = ['LUAD', 'LUSC', 'THCA']
FILTER_CANCER_TYPES = None
filtered_df = results_df[(results_df.signal == 'signal')
& (results_df.data_type == 'test')].copy()
if FILTER_CANCER_TYPES is not None:
filtered_df = filtered_df[(
filtered_df.cancer_type.isin(FILTER_CANCER_TYPES))]
else: