Database Composition
- Training data - Training proteins combined NDRD95, and None-ARG data from UniProt
- Test data - Test proteins combined NDRD, and None-ARG data from UniProt
- NDARO data - From NCBI ARG database
- LARGE Model - Software running models
The dataset is divided into training (64,317 sequences) and testing (1M+ sequences) sets with complete independence between them. This strategic separation enhances model accuracy and reduces false positives.