def vdjdb_beta(q=0, epitopes=False):
vdjdb = parse_vdjdb(vdjdb_location, q=q)
beta = vdjdb[['cdr3.beta', 'antigen.epitope']].dropna().drop_duplicates()
beta.rename(columns={'cdr3.beta':'CDR3',
'antigen.epitope':'Epitope'},
inplace=True)
if epitopes:
return beta
else:
return beta["CDR3"].drop_duplicates()
def vdjdb_paired(q=0, epitopes=False):
vdjdb = parse_vdjdb(vdjdb_location, q=q)
paired = vdjdb[['cdr3.alpha', 'cdr3.beta', 'antigen.epitope']].dropna().drop_duplicates()
paired.rename(columns={'cdr3.alpha':'CDR3_alpha',
'cdr3.beta':'CDR3_beta',
'antigen.epitope':'Epitope'},
inplace=True)
if epitopes:
return paired
else:
return paired[['CDR3_alpha', 'CDR3_beta']].drop_duplicates()
def vdjdb(q=0):
df = parse_vdjdb(vdjdb_location,q=q)
return df[["cdr3.beta", "v.beta", "antigen.epitope"]].dropna().drop_duplicates()
def vdjdb_tcrdist(filename, q=0):
prepared_data = parse_vdjdb(filename, q=q)
prepared_data = prepared_data[['cdr3.beta', 'v.beta']].dropna().drop_duplicates()
prepared_data.rename(columns={'cdr3.beta':'cdr3_b_aa',
'v.beta':'v_b_gene'})
return prepared_data
def vdjdb_gliph2(filename, q=0):
prepared_data = parse_vdjdb(filename, q=q)
prepared_data = prepared_data[['cdr3.beta', 'v.beta']].dropna().drop_duplicates()
prepared_data.rename(columns={'cdr3.beta':'CDR3',
'v.beta':'V'})
return prepared_data
def TCRDist(data=None, chain='beta', sparse=False):
'''
Calculate distances between TCR sequences, using the TCRDist metric
described in Dash et al. 2017 Nature.
Parameters
----------
data : pandas.DataFrame, optional
Input dataframe containing information about
CDR3 sequence, V gene and antigen specificity. The default is None.
chain : str, optional
TCR chain: alpha, beta or paired. The default is 'beta'.
sparse : Bool, optional
Turn on sparse distance computation. The default is False.
Returns
-------
S : numpy.array
TCRDist distance matrix.
seq : pandas.Series
pandas.Series with sequences for which distances have been calculated.
gt : pandas.DataFrame
Ground truth. pandas.DataFrame containing information about the
TCR sequence and its cognate epitope target.
'''
if data is None:
vdjdb = parse_vdjdb(
os.path.abspath('./clustcr/input/vdjdb/vdjdb_full.txt'), q=1)
else:
vdjdb = data
if chain == 'beta':
cdr3 = 'cdr3_b_aa'
v_name = 'v_b_gene'
vdjdb = vdjdb.drop(columns=['cdr3.alpha', 'v.alpha'])
vdjdb = vdjdb.rename(columns={'cdr3.beta': cdr3, 'v.beta': v_name})
elif chain == 'alpha':
cdr3 = 'cdr3_a_aa'
v_name = 'v_a_gene'
vdjdb = vdjdb.drop(columns=['cdr3.beta', 'v.beta'])
vdjdb = vdjdb.rename(columns={'cdr3.alpha': cdr3, 'v.alpha': v_name})
df_epi = vdjdb[[cdr3, v_name,
'antigen.epitope']].dropna().drop_duplicates()
seq = df_epi.drop(
columns=['antigen.epitope']).drop_duplicates().reset_index(drop=True)
gt = df_epi.rename(columns={
cdr3: 'CDR3',
v_name: 'V',
'antigen.epitope': 'Epitope'
})
if sparse:
tr = TCRrep(cell_df=seq,
organism='human',
chains=['beta'],
db_file='alphabeta_gammadelta_db.tsv',
compute_distances=False)
tr.cpus = 2
tr.compute_sparse_rect_distances(radius=200, chunk_size=500)
S = tr.rw_beta
else:
tr = TCRrep(cell_df=seq,
organism='human',
chains=[chain],
db_file='alphabeta_gammadelta_db.tsv',
compute_distances=True)
S = tr.pw_cdr3_b_aa
return S, seq, gt