Training and validation datasets, and pretrained models of VPAC

This repository contains the training and validation datasets, as well as the pretrained models used by VPAC (Virus Prediction from Assembled Contigs). The datasets include simulated viral and non-viral genomic fragments from diverse environments, covering viruses, bacteria, archaea, plasmids, and fungi. They were carefully curated to avoid overlaps with the independent test sets. The pretrained models are derived from large-scale training of VPAC and can be directly used for viral contig classification tasks. The source code of VPAC and its integration within the VMP pipeline can be found at https://github.com/Basspoom/VMP.