ec-knn-rr

CPU-only enzyme EC prediction for the Leak CURBER enzyme classification task.

MMseqs2 search of each unique input against the full training set (prebuilt index). Candidate ECs come from the top hits; a small gradient-boosting re-ranker (when shipped) or a bit-score-weighted vote picks the first EC. The list is best-first. Inputs with no hit get the most frequent training EC. A time guard stops searching early and uses the fallback EC for the remaining (longest) inputs if the machine is slow.

Run: python /app/predict.py --input INPUT.parquet --output OUTPUT.jsonl