YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
DA3410 EC Classification — Retrieval Model
Overview
This repository contains an enzyme EC-number classification model for the Leak CURBER competition. It uses MMseqs2 protein sequence retrieval and weighted voting over the enzyme classification labels of retrieved training sequences.
Required Files
Dockerfile— Defines the Linux inference environment.predict.py— Main prediction entry point.inputs.py— Input validation and preprocessing.task.json— Task configuration.ec_vocabulary.json— Valid EC-number vocabulary.full_retrieval_train_labels.tsv— Training sequence-to-EC-label mappings.mmseqs_db/— Indexed MMseqs2 protein sequence database.
Inference
The evaluation environment runs:
python /app/predict.py --input INPUT --output OUTPUT
The script writes JSONL predictions, with one record per unique input ID. Each record contains an id and a prediction list containing one EC number.
Method
The model retrieves similar protein sequences from the training database and uses weighted voting over their annotated EC numbers to produce a prediction.
Runtime
The inference implementation uses MMseqs2 and is designed to run offline in the competition's container environment.