upset figure from given column

upset.py

DF = pd.read_csv('tmpout/merged.maflite', sep = '\t')

DF.groupby(["pipeline"]).sample_id.nunique()

# # define a "MATCH"
#
# in this section we attempted to use a set of columns to define a matching 

# id_cols defines a unique event
IDCOLS = [
    "contig",
    "startpos",
    "endpos",
    "vtype",
    'altbase', # this is also important to get right
    "cohort",
    "sample_id"
]
unique_DF = DF.groupby(IDCOLS, as_index=False).agg({'pipeline': lambda x: ','.join(set(sorted(x)))})


unique_DF.pipeline.value_counts()

unique_DF.isna().sum()

DF2 = unique_DF.rename(columns = { 'pipeline' : 'called_by' })\
.merge(DF.drop(columns = ['pipeline']), how = 'left')\
.groupby(IDCOLS).first().reset_index()



from upsetplot import from_memberships
from upsetplot import plot

def get_overlap(dfa, title):
  """
  dfa assumes a column "called_by". a comma separated column of multiple pipelines ("washu,getz")
  """
    # upset plot
    dfc = dfa.groupby('called_by').size().reset_index(name = 'count').sort_values('called_by')
    plot(from_memberships(dfc.called_by.str.split(",").tolist(), dfc['count']), 
         sort_by = 'degree', sort_categories_by = None, 
         show_counts = True)
    plt.title(title, y=1.08)
    return(dfa)