def get_other_variants(most_damaging, all_var, AB, Gene, Exon, Date):
''' Get a list of all sample names that contain a given false positive variant
or a variant which does not pass the threshold of the allele balance and
use it to get all other variants asociated with said sample.
Args:
most_damaging: csv containg most damaging variants per sample
all_var: all called variants assocaited with each sample referred to in most_damaging patients
Returns:
A modified all_vars df which has all the alternative variants for each
sample within the most_damaging df except variants below the AB threshold
and within false positives. Returned df is sorted by samle name and variant
score.
'''
# copy and rename columns
df = most_damaging.copy()
df = df.fillna("-")
df = rename.rename_columns(df)
# remove samples that have all NaN entries in the fields of interest
df['all_nan'] = df.apply(lambda x: "Y" if (x['Symbol'] == "-" and x['Exon'] == "-"
and x['AB'] == "-") else "N", axis=1)
df = df[~df['all_nan'].str.contains("Y")]
# convert AB to numeric
df['AB'] = pd.to_numeric(df['AB'], errors='coerce')
# filter for variants with SKI exon1 and AB < 0.3
df = identify_unwanted(df, AB, Gene, Exon, Date)
df = df[df['TEST'].str.contains("LOW", na=False)]
# list of all sample names
l = df['Sample'].tolist()
# copy and rename columns
all_vars = all_var.copy()
all_vars = rename.rename_columns(all_vars)
# filter for only rows that contain sample name in the given list
all_vars['cross'] = all_vars['Sample'].isin(l)
all_vars = all_vars[all_vars['cross'] == True]
# filter for variants with AB > 0.3 or aren't SKI exon 1
all_vars = identify_unwanted(all_vars, AB, Gene, Exon, Date)
all_vars = all_vars[~all_vars['TEST'].str.contains("LOW", na=False)]
# sort in order of sample name and score
all_vars = all_vars.sort_values(['Sample','Score'], ascending=False)
return all_vars
def clean_phenotype_data(phenotype):
''' Clean the phenotype data.
Args:
phenotype: path to phenotype file
'''
unwanted_char = {' ': '', '-': '', '\'': ''}
name_corrections = {
'24SA1565': '21SA1565',
'24SS1575': '21SS1575',
'24GC1574': '21GC1574',
'24DR1571': '21DR1571',
'24FP1566': '21FP1566',
'24GC1574': '21GC1574',
'24AS1570': '21AS1570',
'24KS0915': '24ZS0915',
'1328': '21RL1328',
'24DW932': '24DW0932',
'926': '24GN0926',
'931': '24SG0931',
'937': '24CB0937',
'1327': '21SN1327',
'1374': '24AS1374',
'MK3598': 'MK_35_98'
}
pdf = pd.read_csv(phenotype, encoding='iso-8859-1')
pdf_clean = rename.rename_columns(pdf)
pdf_clean = rename.replace_series_strings(pdf_clean,
'Sample',
unwanted_char,
substring=True)
pdf_clean = rename.replace_series_strings(pdf_clean,
'Sample',
name_corrections,
substring=False)
return pdf_clean
def create_new_most_damaging(old_most_dam, all_vars, AB=0.3, Gene="SKI", Exon="1/7", Date="01-Jul"):
''' Replace the most damaging variant for each patients variant whom
does not pass the allele balance threshold or whoms variant is
within a known false positive gene and exon. If the existing most
damaging variant is the only variant for that patient, then said
variant will remain as the most damaging.
Args:
old_most_dam: existing dataframe which details the most damaging variant for each patient
all_vars: a dataframe which contains all variants associated with the patients detailed in old_most_dam
AB: allele balance minimum threshold
Gene: gene in which a known false positive lies within
Exon: exon of said gene in which a known false positive lies within
Date: converting a xsxl to csv results in exon nums turning into dates i.e. 1/7 becomes 01-Jul. This ensures said exons are filtered if this is the case.
Returns:
The old_most_dam df where the next most damaging variant has been
put in place of the old most damaging variant that did not pass
the allele balance threshold or was within a known false positive
'''
# drop duplicates (dropping is fine as they have been sorted by score)
all_alt_vars = get_other_variants(old_most_dam, all_vars, AB, Gene, Exon, Date)
alt_most_dam = all_alt_vars.drop_duplicates(['Sample'])
alt_most_dam['new_md'] = "Y" # mark sample/variants
# rename columns so they match with alt_most_dam
old_most_dam = rename.rename_columns(old_most_dam)
# append the newly selected most damaging, sort so these variants appear above old most damaging vars and drop duplicates so only new most damaging remain.
append_most_dam = old_most_dam.append(alt_most_dam)
append_most_dam = append_most_dam.sort_values(['Sample', 'new_md'])
new_most_dam = append_most_dam.drop_duplicates('Sample')
return new_most_dam
def get_phenotype_columns(p):
''' Open a phenotype file and return the
column names as a list.
Args:
p - phenotype file
'''
p = rename.rename_columns(pd.read_csv(p, encoding='iso-8859-1'))
phenotype_columns = list(p.columns.values)
phenotype_columns.remove('Sample')
return phenotype_columns
def clean_all_var_df(df, three_categories=True):
''' Clean up of the all variants data. '''
df['Dup'] = df.apply(lambda x: mark_duplicate_samples(x, df), axis=1)
df = df[~df['Dup'].str.contains("Duplicate")]
df = conversion.convert2numeric(df, ['age at diagnosis'])
# rename_columns performed in merge_genotype_phenotype
df = rename.rename_columns(df)
df = rename.rename_entries(df)
df = nc.create_new_columns(df, three_categories)
df = df[~df['Sample'].str.contains("Blank|blank|ddH20|dH2O|H2O|BLANK|ddh2o"
)]
df = sf.no_SKI_exon1(df)
return df
def clean_genotype_data(genotype):
''' Clean the genotype data and filter for the
genotype columns of interest
Args:
genotype: path to genotype file
'''
genotype_columns = [
'Sample', 'AD', 'AB', 'UID', 'validation', 'Category', 'Score',
'Symbol', 'HGVS', 'Chrom', 'Pos', 'Ref', 'Alt', 'Consequence', 'HGVSc',
'HGVSp', 'Exon', 'Intron'
]
gdf = pd.read_csv(genotype, encoding='iso-8859-1')
gdf_clean = rename.rename_columns(gdf)
mask = PLP2VUS(gdf_clean)
gdf_clean.ix[mask, 'Category'] = 'Uncertain Significance'
gdf_clean.ix[mask, 'New Category'] = 'VUS'
gdf_filtered = gdf_clean[genotype_columns]
return gdf_filtered