Novelty-aware and locally adaptable graph learning prioritizes Klebsiella phages for strain-level screening

This repository contains the processed datasets and pretrained models, as described in the manuscript titled "Novelty-aware and locally adaptable graph learning prioritizes Klebsiella phages for strain-level screening". The framework integrates protein sequence, structural, and genomic-context representations to support prediction under host-unseen, phage-unseen, and both-unseen settings, with model routing and fine-tuning for novel entities.