Status: planned. Public source repositories exist, but this repository does not yet contain a public end-to-end fixture for a combined integration.
CLEAN and HIT-EC predict enzyme commission labels from protein sequence representations. Their predictions can complement similarity and domain evidence, but should remain a separate evidence channel with explicit confidence and abstention behavior.
Use immutable releases or commits and follow each project's current model-weight and dependency instructions. Do not treat a mutable clone as a reproducible installation.
Before changing the status from planned:
- add a public or synthetic protein fixture;
- record exact code and model versions;
- verify licenses and model terms;
- define a compact prediction and abstention schema;
- compare the output with independent evidence;
- document expected compute and cache size.
Private sequences, model credentials, and large caches must stay outside this repository.