ec-knn-v1

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

Method: MMseqs2 alignment of each unique input sequence against the full training set, then a bit-score-weighted vote over the ECs of the top hits. The predicted EC set is every EC whose vote is at least half the top EC's vote (at most 3). If a sequence has no hit, the most frequent training EC is predicted.

Files: predict.py (entry point), knn.py (voting logic), bin/mmseqs (static MMseqs2 binary), db/ (training sequences and EC map), config.json (search and voting settings), ec_vocabulary.json (task vocabulary).

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