MetaSRA-pipeline
MetaSRA-pipeline copied to clipboard
Classifier contains outdated CVCL ontology tree
When installing the pipeline it starts with downloading and preparing the used ontologies. This process fetches the newest versions of them which work perfectly fine. However, when running the pipeline like this it frequently raises KeyErrors during the sample type prediction stage. Digging a bit into the code, this is a result of the provided classifier model containing it's own, and possibly outdated, copy of a CVCL ontology tree which may not contain a given key from the newer version and thus fails to retrieve the corresponding terms when trying to hash it. This can be solved however by loading the CVCL ontology and replacing the classifiers copy in the map_sra_to_ontology/run_sample_type_prediction.py
like so:
model.cvcl_og = load_ontology("/path/to/cvcl_obo")