Closed alexanderbonnet closed 1 year ago
Nice catch! Thanks for the report @alexanderbonnet!
I think the API has drifted a little. Ideally we should separate the df_processing_funcs into atom_df_processing_funcs
and hetatom_df_processing_funcs
as the downstream function has entrypoints for both: https://github.com/a-r-j/graphein/blob/28ae687af50cb69f423868c4f49dce968eacea5f/graphein/protein/graphs.py#L238
A quick scan of the logic: https://github.com/a-r-j/graphein/blob/28ae687af50cb69f423868c4f49dce968eacea5f/graphein/protein/graphs.py#L299
Suggests it might not be so robust. Would be very grateful for a PR :)
Now resolved in 1.6.0
pip install graphein=1.6.0
Describe the bug Currently, passing a
List
ofprotein_df_processing_functions
to theProteinGraphConfig
does not have any effect when building the protein graph.To Reproduce
Expected behavior The function should pre-process the
pandas
dataframe before building the protein graph.Desktop (please complete the following information):
pip
.Additional context It looks like the code is missing in the
construct_graph
function here -> https://github.com/a-r-j/graphein/blob/28ae687af50cb69f423868c4f49dce968eacea5f/graphein/protein/graphs.py#L592 Theprotein_df_processing_functions
List
is assigned but never used.Thanks for your work, and let me know if this is something you would like me to take on.