BoostDM | encode_consequence_type(data) inconsinstent when missing consequence.

[FedericaBrando]

boostdm-pipeline/containers_build/boostdm/annotations/utils.py

Lines 7 to 18 in 93d2696

	def encode_consequence_type(data):
	data.loc[~data['csqn_type'].isin(csqn_type_list), 'csqn_type'] = 'none'
	one_hot = pd.get_dummies(data, columns=['csqn_type'], prefix_sep='_')
	one_hot.drop(columns=['csqn_type_none'], inplace=True, errors='ignore')
	for c in csqn_type_list:
	col = f'csqn_type_{c}'
	if col not in list(one_hot.columns):
	one_hot[col] = 0
	return one_hot

pd.get_dummies() in this instance returns by default a dataframe of True and False. However, If a consequence is missing, in the for-loop it is added the missing consequence column filled with 0.

i.e.

SOCS1.DLBCLNOS.annotated.tsv.gz.

csqn_type_missense	csqn_type_nonsense	csqn_type_synonymous	csqn_type_splicing
False			True			False			0
False			True			False			0
False			True			False			0
False			True			False			0
False			True			False			0
False			True			False			0
False			True			False			0
False			True			False			0
True			False			False			0

This inconstistency may derive some problem, in my case I was using a condition like df[df[csqn_col]] expecting boolean column and it failed.

Possible solutions:

1. Uniform all column to 1/0
   • add dtype=float to pd.getdummies() function
2. Uniform all columns to boolean values

   • boostdm-pipeline/containers_build/boostdm/annotations/utils.py

     Line 16 in 93d2696

     one_hot[col] = 0

     
     Replace this line by assigning False instead of 0